By Helga Kolb, Ralph Nelson, Peter Ahnelt, Isabel Ortuo-Lizarn and Nicolas Cuenca

Abstract

We summarize the development, structure, different neural types and neural circuitry in the human fovea. The foveal pit is devoid of rod photoreceptors and of secondary and tertiary neurons, allowing light to directly stimulate cones and give us maximal visual acuity. The circuitry underlying the transmission to the brain occurs at the rim of the fovea. The predominant circuitry is concerned with the private cone to midget bipolar cell and midget ganglion cell pathways. Every cone drives two midget bipolar cells and two midget ganglion cells so that the message from a single cone is provided to the brain as a contrast between lighter signals (ON pathways) or darker signals (OFF pathways). The sharpening of this contrast message is provided by horizontal-cell feedback circuits and, in some pathways by amacrine circuitry. These midget pathways carry a concentric color and spatially opponent message from red and green cones.

Blue cones are sparse, even largely missing in the foveal center while occurring at somewhat higher density than elsewhere in the cone mosaic of the foveal slope. Signals from blue cones have different pathways to ganglion cells. The best understood is through an ON-type blue-cone-selecting bipolar cell to a non-midget, small bistratified ganglion cell. An OFF-center blue midget bipolar is known to be present in the fovea and connects to a blue OFF midget ganglion cell. Another OFF blue message is sent to a giant melanopsin ganglion cell that is present in the foveal rim area, but the circuitry driving that is less certain and possibly involves an intermediate amacrine cell. The H2 horizontal cells are thought to be feedback neurons primarily of the blue cone system.

Amacrine cells of the fovea are mostly small-field and glycinergic. The larger field GABAergic amacrines are present but more typically surround the fovea in a ring of processes, with little or no penetration into the foveal center. Thus, the small field glycinergic amacrines are important in some sort of interplay with the midget bipolarmidget ganglion cell channels. We have anatomical descriptions of their synaptology but only a few have been recorded from physiologically. Both OFF pathway and ON pathway amacrines are present in the fovea.

The central point of the visual field ahead of us is the image falling on the fovea in the human retina. This is the area of our visually sensitive retina where the cone photoreceptors are tightly packed, where rod photoreceptors are excluded and where all intervening layers of the retina are pushed aside concentrically to allow light to reach the densely packed sensory cones with minimum scatter from overlying tissues. The fovea is where focusing on fine detail in the image is perfected, allowing us to read, discriminate colors well and sense three-dimensional depth.

General features of the fovea

Figure 1. The normal human retina fundus photo shows the optic nerve (right), blood vessels and the position of the fovea (center).

Looking at the retina lining the back of the eyeball in a human, we can see the clear landmark of the optic nerve head (papilla) and radiating blood vessels (Figure 1). Temporal to the optic nerve head at a distance approximately 2.5 optic nerve (disk) diameters at roughly 3.4 mm distance lies a dark brown-yellowish area (Figure 1), in the center of which is the tiny circular fovea. The position of the fovea can be seen clearly in the retina illustrated in Figure 2A. This eye was treated with RNA-later for preservation, allowing for a clear view of a yellow macula lutea area and including the brown central point, (foveal pit) (Figure 2A).

Figure 2A.An isolated human retina shows the optic nerve (right), blood vessels and the fovea (center) with surrounding macula lutea (yellow). Cuenca et al, prepublication.

The area called the macula by ophthalmologists is a circular area around the foveal center of approximately 5.5 mm diameter (Figure 2B) The macula lutea with the yellow pigmentation extends across the fovea into the parafoveal region and a little beyond. This area is about 2.5 mm in diameter (Figure 2B). The actual fovea is about 1.5 mm in diameter and the central fovea consists of a foveal pit (umbo) that is a mere 0.15 mm across (Figure 2B). This foveal pit is almost devoid of all layers of the retina beneath the cone photoreceptors. On the edges of the foveal pit the foveal slope is still mainly devoid of other layers but some cell bodies of retinal interneurons, bipolar and horizontal cells and even some amacrine cell processes are becoming evident. By the 0.35 mm diameter circular area the first ganglion cell bodies, the retinal neurons sending signals to the brain, are beginning to appear. All the central fovea that measures 0.5 mm across is avascular (FAZ).

Figure 2B.A map of the whole macular area to show the dimensions of the foveal pit, foveal avascular zone, parafovea, perifovea, and the limits of the macula. Inset shows the dimensions of the foveal avascular zone, which is the fovea we are discussing here.

The avascular nature of the central fovea is depicted in Figure 3. A human retina wholemount has the blood vessels immunostained with antibodies against Collagen IV and is photographed by stacked images in a confocal microscope. It is absolutely clear that the smallest capillaries even, do not intrude into the foveal center (Figure 3, f) of 500 m diameter, thereby known as the avascular zone.

Figure 3.Wholemount of human retina with blood vessels immunostained with Collagen IV. The confocal microscopy of stacked images clearly shows the optic nerve head (ON) and all the blood vessels to the smallest capillaries. The capillaries surround the fovea (f), but do not enter it, thereby making the fovea avascular.

In vertical section of the human retina from the optic nerve head through the foveal pit and beyond (Figure 4), it is clear where the fovea is located relative to the nerve head (on). Figure 4 (a) is a confocal image after immunostaining with antibodies that are specific for cone photoreceptors [arrestin antibodies for cones, green; cytochrome C antibodies for mitochondria, blue; and for Mller glial cells and RPE, antibodies against cytoplasmic retinaldehyde binding protein (CRALBP), red]. In comparison is seen an optical coherence tomography (OCT) picture in Figure 4 (b) of exactly the same area of human retina. In both images it is clear that the second and third order neurons of the inner nuclear and ganglion cell layers respectively are not present in the foveal pit.

Figure 4.(a) An immunostained human retina section covering the optic nerve (ON) and the foveal pit. Cones, anti-arrestin (green); pigment epithelial and Mller cells, anti CRALPB (red) (109); mitochondria, anti-cytochrome C (blue). (b) An OCT image of the same retinal area in a normal human subject. The second and third order neurons of the retinal inner nuclear and ganglion cell layers respectively are not present in the foveal pit. Adapted from Cuenca, Ortuo-Lizarn and Pinilla 2018 (110).

In the foveal pit the only neurons are cone photoreceptors, all with slim inner segments, packed cell bodies, up to 6 layers deep reaching to the floor of the foveal pit (Figure 5, green cells). However, there are many expanded-looking Mller glia surrounding these cones (Figure 5, red profiles). A central bouquet of cones has their synaptic pedicles ending at the foveal pit floor (Figure 5, green spots, arrows), whereas the cones surrounding them stretch their axons (known as Henle fibers) and presynaptic pedicles away from the center of the foveal pit to the foveal slope area (Figure 5, green spots form a continuous line, arrows). The lack of blood vessels in the central pit can be seen by the absence of the blue circular profiles there (Figure 5, bv).

Figure 5.Vertical section of the human fovea immunostained with antibodies to cone arrestin (green), CRALBP (red) and Collagen IV (blue).

OBrien and colleagues (1) very elegantly illustrated the cone axons radiating out from the foveal pit forming the Henle fiber layer and terminating in distant pedicles in a whole mount monkey retina (Figure 6). The picture would be very similar in a human retina. The Henle fiber layer is a combination of outward radially directed axons of the cones, and where rods begin to appear, also rod axons, and Mller cell processes. It is interesting to note in Figure 5 that the pedicles of the very central bouquet of cones are widely spaced ending on the foveal pit floor. We know from Figure 5 that these central bouquet cone pedicles are separated by voluminous Mller cell elements.

Figure 6.A wholemount monkey fovea immunostained with cone arrestin. The axons of the cones radiate out to a ring of cone pedicles. Central bouquet cone axons stay in the foveal pit. From OBrien et al., 2012 (1).

Understanding how the primate fovea develops from fetal to adult stage of the retina has been a very difficult task in vision research. This has, of course been due to the difficulty of obtaining retinas from human pre-birth and baby eyes. Even fetal monkey material has been scarce to obtain. Dr. Anita Hendrickson (Figure 7) at the University of Washington, Seattle, spent most of her career pursuing this subject of retinal research, and has contributed almost all we know.

Figure 7.A young Anita Hendrickson at her microscope. From her obituary in 2017 (111).

The earliest fetal retinas examined (2) were from a week-22 eye. The fovea is not recognizable at this stage, because the central region of the retina, where the fovea will develop, consists primarily of several layers of ganglion cell bodies and inner nuclear layer cells (INL), presumably amacrine and bipolar cells (Figure 8, a). A single layer of developing cones stretches from outer plexiform layer (OPL) to pigment epithelium and choroid (Figure 8, a, right inset). A hint of a developing cone pedicle is seen (Figure 8, right red arrow) but there is no sign of outer segments of cones (Figure 8, right, apposing red arrowheads). By fetal week 28, an indentation of the retina at the thickest ganglion cell layer appears and can be considered the earliest sign of the foveal pit (Figure 8, b, P). The inner nuclear layer has become thinner and appears pushed out of the pit (P) but a kind of split is occurring in the middle of the INL known as the transient layer of Chievitz (TC, Figure 8, c) (3). By fetal week 37 (Figure 8, c) a pronounced foveal pit is evident (P), the ganglion cells are thinned to 2 or 3 deep and the TC area in the INL appears like a sheared, radially projecting area of probable Mller cell fibers. Through the latter two fetal stages, where the foveal pit is becoming obvious, the cones are still immature, arranged in a single layer and have no visible outer segments (Figure 8, b and c). However, there is the first suggestion of the cone axons being tilted away from their cell bodies to form the early Henle fiber layer.

Figure 8.Foetal human retina at (a) foetal week (Fwk) 22, (b) Fwk 28, and (c) Fwk 37. The foveal position is not noticed at week 22 but in later weeks becomes dimpled as ganglion cells become displaced out radially from the developing foveal pit. In the beginning the retina is thick, multilayered and cones are undeveloped with no outer segments or visual pigment (a: right enlarged photo, red arrow heads point to a cone nucleus, a stubby inner segment, and a developing cone pedicle). From Hendrickson et al., 2012 (26).

It is interesting to closely examine the cone photoreceptors in the fetal 35-to-37-week retinas as illustrated by Hendrickson and coauthors (2). Figure 9 shows how immature the cones of the foveal pit are compared with those of the cones at some distance from the fovea (Figure 9. 2 mm from fovea). At the foveal pit area, the cones are just stubby cells with a synaptic pedicle, little to no lengthened inner segment and zero outer segments (Figure 9, fovea). By 800 m to 2 mm from the developing foveal pit, the cones become elongated vertically and have definite cone pedicles. Most cell bodies descend away from the external limiting membrane and have elongating axons that are angled away from the foveal pit, forming the early Henle fiber layer. Inner segments are long, but the outer segments are still not formed. (Figure 9, 800 m and 2 mm).

Figure 9.Sections of the retina of a human foetus at 25 weeks gestation. The cones of the fovea are still undeveloped with no outer segments, and a synaptic area with no axon. From 800 m to 2 mm from the foveal center there are clear elongated inner segments but still no outer segments. The slanting of the cone axons out radially is beginning to be evidence of a developing Henle fiber layer. From Hendrickson et al., 2012 (26).

At birth of the human baby the retina in the eye is looking recognizably foveate (Figure 10, a). The foveal pit now contains a very thin, only one layer thick, ganglion cell layer, a thin inner plexiform layer (IPL) but a prominent inner nuclear layer (INL) (Figure 10, a). The cones are now evident as straight vertical cones with synaptic pedicles, cell bodies and inner segments. There are probably developing cone outer segments too (not easy to see at this magnification). But the pit is still several cell layers thick with only the cones on the foveal slope beginning to angle away from the pit. Further out on the foveal slope the cone Henle fiber layer is obvious now (Figure 10, a). By 15 months after birth, the baby retina has a definite fovea and even the central cones are angling out to the foveal slope. Inner and outer segments are well developed in the pit and no other layers of the retina are here anymore (Figure 10, b and c). By 13 years the fovea is completely developed (Figure 10, d) (2).

Figure 10.The foveal retina sections of a human from (a) postnatal 8 days (P8d), through (b) 15 months, to fully formed (d) 13 years. (c) At 15 months the cones are thin, have outer segments and squash together and, except for the central bouquet, send axons radially outwards as the Henle fiber layer. Second order neurons and ganglion cells are pushed along the foveal slope to form a pile of ganglion cell bodies at the foveal rim. From Hendrickson et al., 2012 (26).

What forces could cause this remarkable transformation of an evenly thick multi-cell, layered retina to become concavely dimpled, buckled up and stretched outwards to form a single layered pit at the fovea and a high sided sloping tissue with the highest concentration of cell layers at the foveal rim. The developmental effort is to ensure that a central area of the retina is concentrated with the slimmest packed cones with no obstruction of incoming light by secondary and tertiary cell layers.

The most recent investigations on this developmental phenomenon in the human (primate) retina provide evidence that the radial retinal glia the Mller cells and possibly the astrocytes of the ganglion cell layer are instrumental in this process (4). The Mller cells of the foveal pit are closely associated with the cone fibers and together they make up the Henle fibers layer (Figure 11A, red profiles). Bringmann and colleagues suggest that the Mller cells exert tractional forces onto cone axons fibers by a vertical contraction of the central most Mller cells and cones so they become elongated and very thin (Figure 11, B, blue arrows). After widening of the foveal pit by elimination of astrocytes in the pit and ganglion cell layers, the Henle fibers are forced, by horizontal contraction of their surrounding Mller cell processes in the outer plexiform layer, to pull the cone and then rod photoreceptor centrifugally away from the pit (Figure 11, B, orange arrows).

Figure 11.(A) A human fovea drawing to show that the Henle fiber layer consists of cone photoreceptor axons as well as envelopingMller cells and fibers (red). B) Drawing to show the central foveal cone bouquet of thin and closely packed cones in the foveal pit. The cone axons on the foveal slope move radially out with the Mller cells to form the Henle fiber layer and end in pedicles that make connection with bipolar cells at some distance from the foveal pit. Blue arrows show the vertical squeezing and packing of the cones in the foveal pit and orange arrows show the displacement horizontally of the foveal cone axons, during development of the adult fovea.

The term foveal cone mosaic generally refers to the strikingly regular patterns of condensed cone inner and outer segments with largely triangular crystalline organization, which nevertheless includes non-randomly distributed discontinuities (5, 6). The less familiar and less understood part of foveal cones is the further course towards their synaptic terminals. It includes a two-step transition. From a two-dimensional mosaic for image reception it is rearranged into to a three-dimensional somata tiling, which then again spreads out to establish the concentric monolayered pedicle meshwork (7-9).

The mature human fovea consists of 3 spectral types of cone: red or long wavelength sensitive cones, L-cones; green or medium wavelength cones, or M-cones; and blue or short wavelength cones, S-cones. These three types of cone are tightly packed and at their most concentrated (up to 200,000/mm2 in the fovea (8, 10) (see Webvision Facts and Figures). Rods are not present in the foveal pit, appearing first halfway into the foveal slope, beyond the 300 m diameter area (see Figure 2B).

It is extremely difficult to get a horizontal section through the central fovea particularly including the central bouquet of cones because of the concave nature of the fovea. Figure 12.1 manages to get such a view of a horizontal slice through the inner segments of the cones of a human fovea (7). The tiniest central cones in the center of the photograph (Figure 12.1) are very slim at 2.5-3 m in diameter and become progressively larger as they move along a radial gradient from the central bouquet. It is noticeable that the cones are not uniformly distributed in a hexagonal mosaic. Small patches of cones are hexagonal and then the patch is interrupted and shifts the surrounding patches slightly (Figure 12.1). Ahnelt and coauthors (11) noticed that these shifts in the mosaic usually were associated with the position of a slightly larger diameter cone. They proposed that these larger cones were the short wavelength cones, the S-cones, and described their morphological differences from the surrounding, more common L- and M-cones (11).

Figure 12.1.A horizontally sectioned and stained human retina at the foveal pit and rod free area. From Ahnelt et al, 1987 (11).

S-cones are relatively rare in the retina compared with the much more dominant L- and M- cones. The S-cones are, however, ubiquitous in all vertebrate retinas, with the exception of cetaceans (12). As far as other mammals are concerned S-cones are commonly paired with L-cones to give them a dichromatic color sense. These L-cones vary in spectral peak, and the more mid-spectral types are called M-cones. In old world monkeys and apes, and in man an L-opsin gene duplication and further mutation produced an extra mid-spectral L-cone opsin subtype, M-cone opsin. The combination of L-cones, M-cones and S-cones provides trichromacy. This trichromacy allows discrimination of green, yellow and blue/purple hues.

There are differences in the genetic structure and locus of the S-cone visual pigment compared with the M- and L-cone pigments (13), yet the S-cones always form a consistent 8-10% of the mammalian cone photoreceptor population (14, 15). In primates and humans of course, the S-cones are rather scarce in the foveal pit. Some authors suggest that there is a so-called blue cone blind spot (16). However, S-cones peak in number on the foveal slope of the human retina and here form about 12% of the population. Figure 12.2, (a) shows the peak S-cone distribution on the foveal slope in a human retina as identified by the larger size and arrangement in the mosaic breaking up the regular hexagonal pattern distribution of the other cone types. In Figure 12.2, (b) the S-cones have been colored in for clarity.

Figure 12.2.A whole-mount photograph of the foveal slope of a human retina. P (upper right corner) is the foveal pit. Larger cone profiles break up the mosaic of cones into disjointed groups of closely packed smaller profile cones [arrows in (a, b) and colored in as S-blue cones in (b)]. From Ahnelt et al., 1987 (11).

Since these earlier identifications of foveal S-cones on morphological criteria (11), antibodies against the S-cone pigments in the cone outer segments have been developed and are able to positively identify the S-cones in the overall population by immunocytochemical methods. In figure 13, the human foveal pit (FP) and foveal slope are immunostained with an S-cone antibody and illustrate the S-cones as black spots and angled black cone outer segments. In the foveal pit only a few S-cones appear interspersed in the mosaic of highest density (Figure 13). However, their proportion increases in surrounding areas and are at their highest density on the foveal slope (Figure 13 brown spots, top and right-hand side).

Figure 13.The foveal pit (FP) and part of the foveal slope are immunostained with an S-cone opsin in a human retina.

Figure 14 illustrates immunostaining in vertical section and the scarcity of S-cones in the foveal pit compared to the increase in number of this population of cones on the foveal slope, of a human retina. A map of the S- cone distribution in another human fovea is shown in Figure 15. The lighter to darker blue shading indicates less dense to denser S- cone presence. Note in both images (Figs. 14 and 15) there are very small numbers of S-cones in the foveal pit.

Figure 14.Vertical section of a human foveal pit immunostained with antibodies against cone arrestin for all cones (red), and JH455, which labels S-cones (green). Few S-cones are found in the foveal pit.

Figure 15.Every S-cone is labelled with S-cone opsin antibody in a human fovea. The more intense blue shading indicates greater densities of S-cones in the foveal slope where they reach 12% of the cone population.

It has been rather easy to identify S-cones in the human fovea and the rest of the retina by these immunocytochemical techniques where S- cones can be visualized and distinguished from the surrounding L- or M-cones. Figure 16 shows a spectacular confocal image of the cones in near peripheral human retina by immunolabeling with cone arrestin, and by the HJ455 antibody to S-cones, that shows up the S-cone opsin both in the outer and inner segments.

Figure 16.Near peripheral retinal human cones stained with HJ455 antibody that identifies the S-cones (green) amongst the arrestin (red) labeled cones.

Sadly, the L-cones and M-cones are not distinguishable on immunostaining techniques because their visual pigments are so close in structure. There is presently no antibody developed to separately mark them into L- or M- cone types. So, to identify L- and M-cones in the human fovea we must go to other more sophisticated techniques. Psychophysical measurements have suggested that L- cones usually outnumber M-cones by 2:1 in the human fovea (17). Microspectrophotometry of all cones in small patches of cones in the fovea of monkeys, has revealed that L- and M-cones occur in about equal proportion (18).

Newer techniques, introduced by Roorda and Williams (19), use adaptive optics to make direct measurements of spectral sensitivity of foveal cones in the living human eye (Figure 17). They found that humans varied greatly in the proportions of L-cones to M-cones: some individuals have almost equal proportions while others have a higher proportion of L-cones, even to the extreme of 16 L-cones to every M-cone (Figure17, BS). While the sparser S-cones are spaced regularly, the L- and M-cones lie randomly in the mosaic meaning that clusters of cones of the same spectral type will occur together as suggested from Mollon and Bowmakers paper (18). Roorda and coauthors (20) concluded that L- and M-cones are in a random distribution in the foveal center (21). Nevertheless, the human subjects HS and BS in Figure 17 would seem intuitively to have a different perception of color. But both subjects were reported to have normal color vision (19). A single cone is achromatic, and its stimulation doesnt result in color vision unless there is comparison to stimulation of a neighbouring cone with different opsin (22). This comparison is done by retinal and brain neural circuitry (see later section on horizontal cell roles in spectral antagonism). Some elegant recent human adaptive optics studies and psychophysical reporting found that 79% of targeted cones in the foveal center, tested for color perception, correctly identified the color (hue) (22). Interestingly, others, using similar techniques of adaptive optics and human reports of hue for single cone stimulation with colored light in the fovea, found a considerable proportion of cones produced only white sensations (21).

Figure 17.Method of adaptive optics shows mosaics of L (red), M (green) and S (blue) cones in four human subjects with normal color vision. The ratio of S to L and M cones is constant, but that of L to M cones varies from 2.7:1 (L:M) to 16.5:1 (L:M). Adapted from Roorda and Williams, 1999 (19).

The process of centrifugal displacement by the Henle layer affects cone pedicles in different ways, depending on their eccentricity (Figure 18).

Figure 18.Foveal pit in blue and the foveal slope to the foveal edge in grey. Cone pedicles lack telodendria in the foveal pit. Pedicles with increasing eccentricity along the slope have tadpole-like shape. More peripherally cone pedicles are round in shape and have telodendrial interconnections. The transition coincides with the appearance of capillaries (red) and microglia (green spots). The thin blue line denotes the elliptical course of the external limiting membrane sectioned at the foveal slope at 1 degree (300 m eccentricity).

In the central bouquet of cones in the foveal pit, the pedicles appear to stay in place (Figure 18). In serial semithin (Figure 19, a) and electron microscopic (Figure 19, b) sections, a few roundish pedicles can be found at the foveal floor (Figure 19, a-c, circles). They are isolated from each other, thus lacking any connections to other cones via telodendria. Still they are contacted by dendritic processes running horizontally from a few interneurons (presumably bipolar and horizontal cells) from the foveal slope or even those neurons lying embedded in voluminous Mller cell processes (Figure 19 b-c, red circles around pedicles).

Figure 19.LM and EM appearances of cone pedicles. (a), (b) and (c) are isolated pedicles of the foveal pit (red circles). There are large Mller-cell processes and neural processes running to the cone pedicles. (d) and (e) show tadpole-like cone pedicles on the foveal slope. (f) Pedicles at the first capillary zone are arranged in curved, bead-like series. (g) Higher magnification shows the telodendrial network between most cone pedicles in (f). (a) is from Ahnelt, 1998 (112), ganglion cell (gc), Mller cell (Mc), cone axon (ax), scale bar 50 m. (g) is from Ahnelt and Pflug 1986 (113).

From the outer central cones, Henle fibers of short length terminate in peculiar tadpole-like pedicles (Figure 18, Figure 19, d-e). They too are largely isolated from neighboring terminals and are characteristic of the cone pedicles until about 1 or 288 m out (23). Beyond this zone still almost entirely established by cone terminals only the pedicles make up a patchy mosaic (Figure 19, f-g). These terminals elaborate telodendrial networks that end on neighboring cone pedicles at gap junction connections (1, 24). This pedicle mosaic tends to establish radial arrays yet is locally influenced by interspersed glia (Figure 19, g).

The cones of the foveal pit project vertically downwards (Figure 20, a). As the concentrated central cones have to extend their axons radially out of the pit they, together with Mller cells, become the Henle fibers. The cone axons become longer and longer as they project onto the foveal slope and into the parafovea (Figure 20, b, 200-400 m long). From then on, further out into the perifovea, the axons begin to shorten and by 3 mm eccentricity from the foveal pit axons are essentially no length at all (Figure 20, c-d, 4000 m periphery). The Henle fiber layer is over as is the macula lutea (Figure 2A, Figure 2B).

Figure 20.Cone morphology in the foveal pit (a), foveal slope (b) and peripheral retina (c). Cones and ON bipolar cells are immunostained with GNB3 (green). Drawing (d) shows the cone morphologies in the different areas. An S-cone (blue-green) is shown in comparison with the M/L-cone types.

S-cones and M/L-cones differ in the time course of mitotic differentiation and expression of opsins. According to Xiao and Hendickson (25), S-opsin and various synaptic proteins are detectable at fetal week 11, while various synaptic and transduction proteins appear in M/L cone subclasses before their opsin visual pigments are detected at fetal week 13 (26). It is clear that S-cones develop in a different mosaic than M/L-cones. Ahnelt and coworkers (7) have noted that cones likely to be short wavelength sensitive tend to occur in irregular positions in both, foveal and peripheral areas. Figure 21A shows an opsin labeled S-cone (asterisk) positioned between seemingly linear series of unlabeled M/L-cone inner segments. Thus in the foveal all-cone mosaic, S-cones appear to interrupt the linear beads of L/M cone-cell inner segments and clearly do not belong to the mosaic of M- and L-cones (6).

Figure 21A.Human cone inner segment mosaic on the foveal slope. Note the first rod (r), and the bead-like arrangement (colored lines) of the M- and L-cones circumventing an S-cone labeled by an S-opsin antibody (asterisk).

The S-cones form a random mosaic like the M/L cones except at the foveal slope area where they are at highest concentration. Here they approach a non-random distribution (25).

Figure 21B shows a schematic summary (7) of cone arrangement in the mosaic of the foveal slope area where the S-cones develop first and reach the non-random mosaic arrangement (25, 27). Three L/M cone patches are exemplified with false colors (yellow, dark blue green and light green). These have migrated downward from an initial position near the external limiting membrane (ELM) to form bead-like arrangements of M/L cone cell bodies in the depths of the outer nuclear layer (ONL). Their axons (Henle fibers) emerge from the cone nuclear layer and radiate centrifugally towards their pedicles. At the intersection of the L/M patches sits an S-cone always with its cell body, unmigrated, up at the outer limiting membrane. Figure 21B left top, indicates the original position (transparent ovals) of M/L cell bodies before mosaic condensation and their presumed path (tapered rays) to their adult positions.

Figure 21B.The transformation of the foveal cone mosaic groups (yellow, dark green, light green) by condensation of their inner/outer segments to vertical sequences of beaded cell bodies and descending, radiating axons in the Henle fiber layer. At left, the original position of the yellow groups cell bodies (line of ovals) before mosaic condensation is indicated, as well as their eventual path (curved lines) to their adult positions. Apparently, S-cones (blue) do not participate in this process, as their cell bodies stay close to the ELM (external limiting membrane, large arrow). Adapted from Ahnelt et al, 2004 (7).

As we have illustrated in Figure 2B, the whole fovea is roughly 1.5 mm across and so any cell found within 750 m of the foveal center is considered a foveal associated cell. It has been hard to get good staining of horizontal cells (HC) of the fovea but some Golgi impregnated human retinas in our possession did allow us to see a few within the 750 m of eccentricity around the central foveal pit (Figure 22) (28).

Figure 22.The shape and size of horizontal cells in the human fovea (Golgi staining). The smallest HCs are in the avascular zone edge of the foveal slope (350 m). The closest HCs stained on the inner foveal slope (200 m) are stretched out, with dendrites following the circular foveal pit circumference and reaching into the central bouquet of cones. From Kolb et al., 1994 (28).

The closest to the foveal center, which is of course cell free except for cone photoreceptors and some dendrites running up to synapse with the central cones, would be the HC at 200 m from the foveal center (Figure 22, top cell). These horizontal cells are elongated and arranged concentrically in a circle around the foveal center and on the far edge of the foveal pit. The area could still be in the avascular zone. Note the dendrites are reaching quite far to contact central cones. The cells are axon bearing, but morphologically it is difficult to judge of which type. The cells at 350 m (Figure 22) are much smaller than the foveal edge HC but now recognizable as H1, H2 and H3 cell types (28). The smallest are the H1 cells that appear to contact about 4-5 cones, judging by their dendritic clusters. H2 cells are wirier and more irregular than H1 and H3 cells but have quite closely packed and profuse dendrites (Figure 22). These H2 cells would be reaching into the foveal slope area, where we know there is the highest density of S-cones, to contact the latter cone type. H3 cells may also be reaching into the foveal slope but we know from previous data they do not receive synapses from S-cones (29, 30). There are no evident axons on these Golgi stained horizontal cells (Figure 22, 350 m), which probably reflects understaining.

The three horizontal cells at 500 m from the foveal center (Figure 22) would also be foveal HCs but in an area where blood vessels occur and the first rod photoreceptors are present. As can be seen they are a little larger in dendritic field size (Figure 22). The H1 cell contacts 6 cones and the H3 about 8-9 cones (Figure 22). H1 and H2 types here have axons (small arrows in Figure 22), which will expand into axon terminals in contact with rods in the case of H1, and with S-cones in the case of H2 cells (31).

By confocal microscopy the central human fovea can be seen to contain parvalbumin immunoreactive horizontal cells (Figure 23, a-b; green cells under the cone pedicles). Parvalbumin identifies H1/H3 horizontal cell types and it is likely that the Golgi staining at the 200 m distance from the central foveal pit is therefore of these types. They are elongated and not closely packed. Their dendrites would be reaching to contact central foveal bouquet cones (Figure 23, b). In contrast, the H1s of the foveal slope are closely packed with vertically squashed cell bodies and small bushy dendrites reaching to the closely packed cone pedicles at the ends of the Henle-fiber-layer cone axons (Figure 23, c). These HCs are clearly the same as those in the Golgi preparations at 300-500 m (Figure 22).

Figure 23.Vertical section of the human fovea cut along the edge of the foveal pit. H1 horizontal cells are immunostained with anti-parvalbumin (green) and cone photoreceptors with recoverin (red). H1 cells are very crowded together in the foveal slope.

The H2 cells of the human retina are known to be particularly associated with the S-cone (blue) photoreceptors (see Webvision chapter on S-cone pathways). We know that H2 cells stain with antibodies to calbindin in the human retina as compared to parvalbumen staining for H1/H3 cells. Figure 24 (white arrows) shows a few calbindin positive HCs (red cells, arrows) on the foveal slope in human retina. In addition to the H2 cells with cell bodies close to the OPL, there are diffuse cone bipolar cells contacting several cones, and amacrine cells stained with calbindin. These red, diffuse bipolar cells have cell bodies lower in the inner nuclear layer and long slanted single apical dendrites as compared to the red H2 cells. Note in this section of human fovea the first rods are present on the foveal slope and the first rod bipolar cells are staining for the antibody to PKC (Figure 24, green cells).

Figure 24.Human foveal slope area immunolabeled with antibodies against calbindin (red) that marks H2 horizontal cells, some bipolar and some amacrine cell types. H2 cells are marked with arrows. The first rod bipolar cells on the foveal slope are labeled with PKC-alpha antibodies (green).

Horizontal cells of the vertebrate retina are known to have important roles in sharpening and scaling of responses from photoreceptors through the subsequent retinal pathways to influence the ganglion cell output (32). At the first level of the outer plexiform layer, horizontal cells are involved in feedback of signal from surrounding cones to each individual cones receptive field. This surround input is expanded well beyond the horizontal cells dendritic connectivity field by virtue of gap junctions that join the dendrites of many horizontal cells of the same type together. i.e. in human retina the H1-H1 cells would be joined in gap junctions and the H2 cells would likewise be joined to other H2 cells (See the Webvision chapter Myriad roles for gap junctions in retinal circuits). This large feedback effect provokes an expanded region of antagonistic signal compared with the central cone signal. In the case of M- or L-cones the antagonistic surround is a mixed M- and L-cone signal. In other words, individual M- and L-cones do not show classic spectral opponency just mixed M- / L-cone surround antagonism (33). The feedback in the case of an S-cone would come from H2 cells, whose contacts include surrounding M- and L-cones. Indeed S-cones have been recorded from in monkey retina and found to have blueyellow spectral opponency as well as center-surround organization (34, 35). Presumably spatial opponency would be transmitted from the M- and L-cones to their respective bipolar cell connections, and in the case of the S-cone, a true spectral opponency has been proven to be transmitted as well (34). No recordings have been made in foveal cones to really see if an M- or L-cone has a spectrally opponent surround like that of (albeit peripheral) S-cones (35).

A long time ago the great Spanish anatomist, Santiago Ramn y Cajal described the neurons of the different vertebrate retinas as seen by sectioned Golgi-stained material. He noted many different types of bipolar cells in the various species and that there were particularly tiny dendritic spreads for some bipolar cells in the bird retina (36). He suggested that these bipolar cells contacted single cones.

In 1941, Stephen Polyak (Figure 25) published books on the neural cell types revealed by Golgi and other silver methods in monkey and human retinas and brain. In central monkey and human retinas Polyak observed and illustrated several types of bipolar cells, but he was very concentrated on the remarkably small dendritic tops of some types that he construed as contacting single cones. He named these bipolar cells, midget bipolar cells (mbc).

Figure 25.Steven Polyak circa 1940.

Figure 26 shows Polyaks original drawing of these midget bipolar cells and larger dendritic field size bipolar cells that would appear to contact several cones (Figure 26, imb, fmb and dfb). Polyak also drew and commented briefly that the midget bipolar cells appeared to be of two varieties, one that had a long axon to the inner plexiform layer, and the other a much shorter axon ending higher in the inner plexiform layer. At the same time, there were midget ganglion cells that had small dendritic trees that came in the two varieties possibly reaching to the axon terminals of the two types of midget bipolar cells (Figure 26, mgcs).

Figure 26.Original drawings of Polyak (90). Bipolar cells and ganglion cells of the central retina. We now know that the invaginating midget bipolar cells (imb) and flat midget bipolar cells (fmb) are physiologically different. Polyak described midget ganglion cells (mgc) as of two types, which we now know are OFF mgc and ON mgc. These connect to fmbs and imbs respectively. Large field bipolar cells (dfb) and parasol ganglion cells were also described by Polyak. The cone spectral types have been colored in by the present authors.

See the original post:
The Architecture of the Human Fovea Webvision