Astrocytes and Blood Vessels Define the Foveal Rim during Primate Retinal Development

Abstract

purpose. To investigate the relationship between development of the perifoveal blood vessels and formation of the foveal depression.

methods. Retinal sections and flatmounts from monkeys aged between fetal day (Fd)80 and 2 years of age were double labeled using antisera to CD31 or von Willebrand factor to detect vascular endothelial cells and antiserum to glial fibrillary acidic protein to detect astrocytes. Sections were studied by fluorescence or confocal microscopy.

results. From Fd88 to 115, vessels on the horizontal meridian were found only at the level of the ganglion cell layer (GCL)–inner plexiform layer (IPL) border where they form the ganglion cell layer plexus (GCP). Stellate astrocytes accompany GCP vessels and extend closer to the fovea than vessels. The foveal avascular zone was present within the GCP at Fd101, and at Fd105 a shallow foveal depression encircled by the GCP was present. The GCP foveal margin had the same dimensions as the adult foveal pit. Both blood vessels and astrocytes were excluded from the emerging fovea throughout development. After Fd140, capillary plexuses in the outer retina anastomosed with the GCP on the foveal slope to form a perifoveal plexus, but this plexus did not mature until a month or more after birth. After Fd142, astrocytes rapidly disappeared from the GCP and most of central retina.

conclusions. An avascular area is outlined by the GCP before the foveal pit begins to form, suggesting that molecular factors in this region exclude both vessels and astrocytes. These factors may also guide neuronal migration to form the pit. Because the perifoveal plexus is formed during late gestation, both capillary growth and foveal development may be affected adversely by prematurity.

The retinas of humans and most monkeys include a central region called the fovea, which is specialized for optimal resolving power. Lateral displacement of inner retinal neurons and glia occurs during late fetal and early infant ages to create the foveal depression. ^{1 2 3 4} The fovea is characterized by a high density of cone photoreceptors centered on the depression, ^{5 6 7} a unique neuronal circuitry, ^{8 9 10 11 12 13} and the absence of a local retinal blood supply. ^{14 15 16 17 18} Several studies have documented changes in the primate retina associated with formation of the foveal depression ^{1 2 3} as well as development of the primate retinal vasculature. ^{16 18 19 20 21 22 23} However, little attention has been paid to changes in the retina that precede formation of the foveal depression, including the temporal relationship between the development of the perifoveal vasculature and the foveal depression (reviewed in Reference 4). 

The location of the future fovea is evident at 11 weeks’ gestation (WG) in the human, ²⁴ or fetal day (Fd) 55 in the macaque, ²⁵ when a region of the outer nuclear layer containing only cones is discernible at the posterior pole. Beginning at midgestation and continuing well after birth, cones are displaced centrally ² to establish adult foveal cone densities of 200,000 to 300,000/mm², which provide the anatomic substrate for the high-resolving power of the fovea. ^{2 26 27 28} The foveal depression is not detectable until approximately 25 WG in the human and Fd110 in the macaque. ^{1 3 4 25 28 29 30} Studies in primates have found that the foveal region is never vascularized during normal development, ^{19 21} but suggest that there is a close temporal relationship between formation of the perifoveal vasculature and the early stages of formation of the foveal depression. ^{4 18 21} It is not known whether the fovea forms before a foveal avascular zone is established or within an already prescribed avascular region. Clarification of this temporal relationship may help unravel the mechanisms directing and controlling foveal development. 

In mammals, retinal blood vessels are first evident in the inner retina surrounding the optic disc and later grow toward the peripheral retinal margin. ^{31 32} This disc-to-periphery maturation of blood vessels contrasts with the fovea-to-periphery maturation sequence of retinal neurons. ^{2 4 29 33 34 35} The primary retinal blood vessels are also associated with astrocytes that accompany and lead the developing blood vessels. ^{18 21 23 36 37} Astrocytes are thought to sense relative hypoxia in the more peripheral avascular regions and, in response, to stimulate endothelial proliferation through the expression of vascular endothelial growth factor. ^{18 38 39} Formation of capillary beds in the outer retina as far as the outer border of the inner nuclear layer (INL) occurs by budding from the primary blood vessels ^{16 31 32} and is thought to be stimulated by vascular endothelial growth factor released from the Müller cells. ^{38 40}  

In primates, the first blood vessels are in the nerve fiber layer plexus that forms at the nerve fiber layer–ganglion cell layer (GCL) interface and initially is distributed in four lobes, one for each retinal quadrant. Vessel growth in the two temporal lobes mimics the arcuate fiber pattern created by ganglion cell axons that pass around rather than through the future foveal region. ^{16 18 20 41} Other retinal plexuses are formed by budding of the nerve fiber layer plexus. In the macaque at Fd120 the inner capillary plexus (ICP) is present at the inner plexiform layer (IPL)–INL border near the optic disc. By Fd130 the outer capillary plexus (OCP) at the INL–outer plexiform layer (OPL) border has also formed near the disc. ¹⁶ However, there are some indications that perifoveal blood vessel development may be different. Growth of human blood vessels toward the foveal region along the horizontal meridian has been described as “retarded,” ¹⁹ and a recent study of the nerve fiber layer plexus development in 14- to 23-WG fetal human retina found very low levels of cell proliferation along the horizontal meridian. ²³  

In the present study we investigated growth of the retinal vasculature around the fovea in fetal and newborn macaque monkeys to establish the temporal relationship between definition of the foveal avascular zone and formation of the foveal depression. Our studies found a distinctive pattern of blood vessel development in this region. The nerve fiber layer plexus was absent, and instead a ganglion cell layer plexus (GCP) was established at the GCL-IPL boundary, from which the ICP and OCP developed by budding. A perifoveal ring formed by the GCP defined the future depression before cell migrations begin within the inner retina. Near birth, the GCP, ICP, and OCP anastomosed to form a perifoveal capillary plexus surrounding the foveal depression. 

Methods

Macaca fascicularis or M. nemestrina monkey fetuses of known gestational age were delivered by cesarean section under halothane or barbiturate anesthesia. Gestation of the macaque is 170 days. Fetuses and postnatal animals were administered a lethal, intravascular overdose of barbiturate before the eyes were enucleated. Eyes were opened at the limbus and immersed in 4% paraformaldehyde overnight. For sections, eyes were cryoprotected and frozen sectioned at 10 μm, parallel to the horizontal meridian. Every 10th section was stained with cresyl violet to locate the fovea and optic disc. Only sections including the optic disc and the foveal cone mosaic were used in this study which included sections from 21 animals in the age range Fd88 to 11 postnatal (P) weeks. The central retina at Fd105, 142, 155, P4 months, and 2.1 years was dissected from the sclera and retinal pigment epithelium (RPE) and stained immunocytochemically as a wholemount. All animal protocols were approved by the University of Washington Animal Care Committee and were in accord with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Immunohistochemistry

Adjacent sections through the foveal region and optic disc were immunolabeled in single or double-label combinations. Astrocytes were labeled using anti-glial fibrillary acidic protein (GFAP; rabbit anti-bovine at 1:500; Dako, Carpinteria, CA). Müller cells were labeled using antibody to cellular retinaldehyde-binding protein (CRALBP; rabbit anti-bovine at 1:20,000, gift of John Saari) and endothelial cells were labeled using mouse anti-human antibodies to the vascular differentiation markers CD31 (at 1:50 to 1:100; Dako) or von Willebrand factor (at 1:50) . 

Sections were incubated for 1 hour in Tris-buffered saline containing 0.2% Triton (TBS-Triton) and 10% goat serum, then incubated overnight in a primary antiserum for single labeling or a mixture of a polyclonal and a monoclonal primary antiserum for double labeling. For single labeling, sections were sequentially incubated for 45 minutes at 37^o in species-specific biotinylated IgG (1:100) followed by avidin-Texas red (Molecular Probes, Eugene, OR; 1:1000). For double labeling, sections were incubated sequentially in anti-mouse biotinylated IgG followed by a mixture of avidin-Texas red (1:1000) and anti-rabbit IgG tagged with fluorescein isothiocyanate (FITC; 1:100). Sections were then washed and coverslipped in 80% glycerol in phosphate buffer. 

Flatmounts of the retina were double labeled to show the distribution of immunoreactive (IR) GFAP astrocytes and CD31-IR blood vessels. Blood vessels in wholemounts at Fd155 or older were labeled with a mixture of CD31 and von Willebrand factor antisera because of a decrease in CD31-IR with age. Whole retinas were rinsed in filtered TBS (pH 7.6) containing 0.002% sodium azide overnight at 4°C, blocked in TBS containing 0.4% saponin and 10% donkey serum for 3 to 4 days, and then incubated in GFAP antiserum in TBS–2% donkey serum at 4°C for 3 to 5 days, thoroughly rinsed in TBS, and incubated in anti-rabbit IgG conjugated to Cy2 (1:100; Amersham, Arlington Heights, IL) for 24 hours at 4°C. Retinas were washed in TBS containing 0.2% saponin and 10% donkey serum for 24 hours, incubated at 4°C in one or a mixture of both blood vessel antisera in TBS–2% donkey serum for 3 to 4 days, washed for 3 hours in TBS, then incubated sequentially in biotinylated anti-mouse IgG (1:50) for 24 to 36 hours and streptavidin conjugated Cy3 (1:100, Amersham) for 30 minutes. Retinas were finally rinsed in TBS and mounted in glycerol with the inner retina uppermost. 

Microscopy

Controls for section staining consisted of the elimination of either one primary or one secondary IgG. There was no bleed-through using the Texas red–FITC filter set used for conventional microscopy and photography. All material presented in this study was judged to be specific, compared with control sections. Double labeling by conventional microscopy was imaged by single-exposure photography for each label, digital scanning of the images, and combination into a single image using the red and green channels of a document generated by computer (Photoshop; Adobe, San Jose, CA). 

Some sections and all flatmounts were imaged in a confocal microscope (Leica, Deerfield, IL) using an argon krypton laser and associated software (TCSNT; Leica). To prepare montages, flatmounts were optically sectioned using a ×16 objective lens in 12 to 19 planes parallel to the retinal surface at 5- to 10-μm intervals beginning at the nerve fiber layer and ending at the outermost capillary plexus. Optical sections were viewed in separate layers using a see-through mode in the image analysis software (Photoshop; Adobe). For analysis and figures, optical sections through the GCP and through the ICP and OCP were merged into single layers. 

Results

General Staining

Blood vessels were intensely IR for the differentiation marker CD31 ⁴² until late gestation, but this declined significantly during the postnatal period. A combination of CD31 and von Willebrand factor antisera showed well-labeled vessels in late-gestation and postnatal retina. Section and flatmount labeling for GFAP detected two morphologic types of astrocytes previously described in monkey retina. ^{16 21 43} Bipolar astrocytes predominated in the nerve fiber layer and had small cell bodies with processes elongated along ganglion cell axon bundles. Large, basophilic stellate astrocytes with long, thick processes were more common in the GCL. Müller cells labeled heavily for CRALBP throughout the period covered by this study but stellate astrocytes never labeled for CRALBP at any age (Fig. 3A 3C) . Astrocytes labeled for GFAP throughout development (Fig. 3A 3B) , but within the foveal region Müller cells also became increasingly GFAP-IR as the foveal depression formed. At all ages Müller cell processes could be distinguished from astrocytes because of their marked vertical orientation which extended through all retinal layers. Stellate astrocytes had large cell bodies and thick processes confined to the GCL. These processes ran horizontally as well as radially and could be distinguished easily from the fine vertical Müller cell processes and punctate foot plates in inner retina (Figs. 3A 3B 3C) . 

Formation of the Ganglion Cell Layer Plexus and Foveal Avascular Zone: Fd88 through 105

Both astrocytes and blood vessels were present on the horizontal meridian at Fd88, extending 1500 to 1750 μm from the optic disc, or approximately halfway to the incipient fovea (Figs. 1A 1B ). Astrocytes were distributed densely throughout the inner retina up to 1 mm temporal to the optic disc (Fig. 1A , left), but beyond this point (asterisk) they were relatively scarce in the nerve fiber layer. Near the disc, blood vessels occurred throughout the nerve fiber layer and GCL (Fig. 1B , left). Between 500 μm and 1 mm from the disc, blood vessels grew across the GCL, so that beyond this point a new plexus was formed at the GCL-IPL border (Fig. 1B 1C 1D 1E 1F 1G 1H 1I 1J 1K 1L 1M 1N 1O 1P 1Q 1R 1S 1T 1U 1V 1W ard the right). This plexus was not found elsewhere in the fetal retina and is called the GCP. Large, stellate astrocytes accompanied the blood vessels and extend 200 to 300 μm in advance of the GCP toward the incipient fovea (compare Fig. 1A , arrow, with Fig. 1B , arrowhead). 

By Fd101 the GCP was present on both the temporal and nasal margins of the incipient fovea (Figs. 1C 1D) . Astrocytes still led the blood vessels by 100 to 200 μm (compare Fig. 1C , arrowheads, with Fig. 2D , arrows). No significant thinning of the GCL was discernible in the avascular region (bracket, Figs. 1C 1D ), suggesting that the inner retinal neuronal migrations that formed the foveal pit had not begun at Fd101. 

In Fd105 sections, astrocytes and blood vessels in the GCP completely encircled the early foveal depression with astrocytes no more than 50μ m farther into the fovea (Fig. 2A , arrowheads). There was a close correspondence between the rim of the depression and the centralmost blood vessel tips and related astrocytes. The inner processes of Müller cells within the avascular area increased in GFAP-IR at this age and this staining increased toward birth as the pit deepened. 

A flatmount of the fellow Fd105 retina shows a well-developed GCP 375 to 425 μm wide encircling the avascular depression (Fig. 2B) . This ring was formed by the anastomoses of four to five large radial vessels feeding into the GCP and had a denser and tighter capillary meshwork than vessels in adjacent retina (Fig. 2C) . Many large, stellate astrocytes were in contact with the perifoveal capillaries, and the larger blood vessels (Figs. 2B , arrowhead, and 2D small arrows), but their overall distribution did not mimic the pattern of the smaller blood vessels. Fd105 appeared to be the peak of astrocyte density in the perifoveal GCP. Astrocytes formed an almost complete circle just central to the perifoveal capillary ring, extending their processes approximately 50 μm onto the slope of the developing fovea. Short, blind capillary branches reached into the fovea as far as this circle, and astrocyte cell bodies seemed to cluster at these branches (Figs. 2B 2D , arrowhead). 

Formation of the Outer Retinal Vascular Plexuses: Fd115 through 145

The outer retina was avascular in sections or a flatmount up to Fd110. In sections at Fd115 near the optic disc, branches of the GCP had grown across the IPL to form the ICP (Fig. 3D , small arrows). A few ICP branches crossed the INL (Fig. 3D arrowheads), indicating that formation of the OCP also had begun. In contrast, near the fovea, blood vessels only were starting to penetrate the IPL (Fig. 3E , small arrows). All GCP blood vessel sprouts were free of GFAP-IR astrocytes. By Fd140, near the optic disc, the four capillary layers were established (Fig. 3F) , but as late as Fd132 only a single layer of GCP astrocytes and blood vessels surrounded the emerging fovea (Figs. 3G ; 4A 4B ). 

In contrast to the progressive inner to outer sequence of plexus formation seen near the optic disc, in the immediate vicinity of the fovea GCP capillary sprouts crossed the IPL and INL and then branched horizontally at the INL-OPL border to establish the OCP as the first plexus in the outer retina (Figs. 3H 4D 4F) . These sprouts were often angled or ran obliquely, were not accompanied by GFAP-IR processes, and had no obvious association with the vertically oriented Müller cell processes. Anastomoses between the GCP, ICP, and OCP were established in the perifoveal retina before the OCP had developed on the foveal rim (Figs. 4D 4F 4H) . 

A flatmount at Fd142 showed that a month before birth, the GCP was well established (Fig. 5A ), but the OCP was only partially formed near the fovea (Fig. 5B) . Anastomoses between the two capillary layers (Fig. 5B , small arrows) were present 150 to 250 μm from the perifoveal ring, but not on the foveal rim. At Fd142 the GCP perifoveal capillary bed was more dense than at Fd105 with many short, blind-ending vascular sprouts reaching toward the foveal depression (Fig. 5C , thick arrows). The number of GCP astrocytes was much reduced (Figs. 4C 5C) , but Müller cells were more heavily labeled. The overall dimensions of the avascular area were 500 × 300 μm, similar to Fd105, suggesting that there had been no further growth of GCP capillaries into the developing fovea. 

Formation of the Perifoveal Capillary Plexus: Fd155 through P3 Week

The precise timing of anastomosis between the ICP, OCP, and GCP to form the perifoveal plexus was found to vary between individuals. Approximately 2 weeks before birth, one Fd155 flatmount showed many anastomoses (Fig. 5D , white arrows) between the outer plexuses (purple) and GCP (red) 100 to 200 μm from the avascular pit. Sections from an Fd160 monkey also demonstrated anastomosis at the level of the ICP between all three plexuses on the foveal slope (Fig. 3F , small arrows). In contrast, sections from a P-1-day animal show that the GCP and OCP were present, but the ICP was minimal, and anastomoses were not present on the foveal rim (Fig. 4F , arrows). 

By P3 week, both the perifoveal plexus and the foveal pit appeared mature and anastomoses were complete (Fig. 4G 4H) . The mature perifoveal plexus was formed by three strata of GCP, ICP, and OCP capillaries with the final step occurring when the GCP and OCP anastomosed with the ICP to form a single layer of vessels at the level of the ICP. The completed perifoveal plexus was seen clearly in a P-2-year flatmount (Fig. 5E 5F) where a single layer of capillaries encircled the avascular fovea. Within 100 μm of the avascular area, branch points could be identified that connected the capillaries of the GCP-ICP (Fig. 5E , white arrows) and OCP (Fig. 5F , white arrows). 

Changing Relationship of Astrocytes to the Perifoveal Capillaries

In all specimens Fd130 or younger, astrocytes lay closer to the fovea than blood vessels in both the GCP and the perifoveal ring (Figs. 1 and 2) , as seen in other parts of the retina. ^{18 21} However, by Fd132 to 140 astrocytes lagged behind blood vessels in the central retina (Figs. 4A 4B 4C 4D) , and there were many fewer astrocytes associated with the perifoveal GCP (compare Fig. 2B with Fig. 5C ). In the Fd142 flatmount, astrocytes no longer clustered around the perifoveal capillaries (Fig. 5C , large arrows). By birth, the immediate vicinity of the fovea was virtually astrocyte free (Figs. 4E 4G , arrowheads), and at P2 year, only eight stellate astrocytes were present in an 800 × 800-μm field (Fig. 5E , green cells). During late gestation and shortly after birth, foveal Müller cells labeled heavily for GFAP, but by P2 year GFAP-IR had disappeared (Fig. 5E 5F) . 

Discussion

The present findings confirm those in previous studies indicating that the immature fovea of human and monkey retina remains free of blood vessels throughout development. ^{16 18 19} In addition, in our study astrocytes also were excluded from the incipient fovea during development, and the avascular region was defined before the foveal depression began to form. Taken together, these findings suggest that during development molecular cues are expressed in the region of the future foveal depression and that these cues include one or more inhibitory factor(s), which exclude vessels, astrocytes, and possibly neurons from the incipient fovea. We also demonstrate that formation of the mature perifoveal plexus involves a unique interaction between the GCP, ICP, and OCP that does not occur in more peripheral retina. 

Growth and Guidance of Retinal Vessels

An early vasculogenesis hypothesis proposed that vascular precursor cells become aligned and then differentiate to form nerve fiber layer capillaries. ⁴¹ Recent studies using endothelial cell–specific markers have failed to label these cells in primate retina. ²⁰ Rather, elongated cells in front of the growing vasculature have been found to be GFAP-IR, indicating that many are astrocytes. Double-label experiments have found that all cell proliferation associated with human inner retinal vascular development can be accounted for within astrocyte and endothelial cell populations, ²³ suggesting that if present, vascular precursor cells do not proliferate. Gariano et al. ²⁰ conclude that almost all the vessel growth in the monkey nerve fiber layer occurs by budding of existing endothelial cells and that this may be under the control of factors released from the astrocytes. This is supported by findings in human retina showing that astrocytes at the actively growing capillary front express vascular endothelial growth factor mRNA. ¹⁸  

Almost all researchers agree that the extension of vessels into the outer layers of the retina involves budding of the existing vasculature. ^{16 18 20 32} Müller cells are suggested to be the main source of vascular endothelial growth factor during growth of vessels toward the outer retina in the rat, ³⁸ but to date this has not been demonstrated in the primate. In the cat it has been reported that the outer retinal plexus forms first in the region of the area centralis, ³² the homologue of the primate fovea. However, in agreement with previous findings in the primate, the present study clearly demonstrates that first the ICP and later the OCP formed first near the optic disc, and then gradually spread toward the fovea, where the full complement of perifoveal capillaries was not established until between Fd160 and P day 1. We also found that in the foveal region, the OCP formed earlier than the ICP, which may be a response to increased metabolic demand due to elevating foveal cone density around birth. ²⁷ It also has been suggested that in the cat the early development of the OCP may be driven by the metabolic needs of central photoreceptors that are born earlier and may become active ahead of those in more peripheral locations. ³²  

Although the migration pathways of astrocytes can clearly influence the growth patterns of retinal vessels, the factors that induce astrocytes into the retina and direct them to migrate toward the periphery are less clear. In general, retinal astrocytes grow in a pattern that reflects the arrangement of ganglion cell axons, ^{18 20 40 44} suggesting that axon bundles provide some mechanical guidance or growth factor stimulation. In primate nerve fiber layer, bipolar astrocytes and vessels migrate in a plane different from that of axons, ²⁰ suggesting that direct contact with axons is not required for either astrocyte or blood vessel growth. In the present study, astrocytes preceded vessels as they grew first across the GCL to establish the GCP and later along the GCL-IPL border. There is no known axonal pathway that might guide this migration, and there is no evidence of involvement of any other type of neuronal or nonneuronal cell. The present findings indicate, therefore, that at least some astrocytes do not require an underlying axon guidance pathway and that factors other than simple cell–cell interactions underlie establishment of the GCP. 

Defining the Foveal Avascular Region

Much recent work in retinal angiogenesis has emphasized the dynamic role astrocytes play in normal and abnormal retinal vascular development, including the secretion of growth factors in response to relative hypoxia. ^{18 38 39 45} Despite the high density of neurons in central retina, which presumably creates a high metabolic demand and in turn should stimulate vessel growth, our present results show that both vessels and astrocytes were inhibited from growing into the incipient fovea. Although human fetal astrocytes were shown to express vascular endothelial growth factor in the nerve fiber layer, ¹⁸ because that study could not be extended after midgestation to allow examination of fovea development, the role of stimulating factors in the development of the ICP, OCP, and perifoveal capillaries remains to be explored. 

The role of inhibitory vascular factors in the retina is poorly understood, ^{22 46 47} but such factors are likely to have a critical role in modulating the action of vasoproliferative factors to control vessel growth. We have shown recently that cell proliferation in the retinal vasculature is significantly reduced along the horizontal meridian of the human fetal retina, including the vicinity of the developing fovea. ²³ This occurs despite the increased number of neurons in the foveal region ⁴ and the probable poor oxygenation of the inner retina by the immature fetal choriocapillaris ¹⁶ that could be expected to enhance endothelial cell proliferation. Both endothelial cells and astrocytes proliferate in fetal retina, ²³ and studies are in progress to determine whether proliferation rates are equally affected. Those findings imply that a factor that inhibits cell proliferation is expressed along the horizontal meridian. In addition, the present finding of blind-ending capillaries that are directed toward but never grow into the foveal region is another indication of the presence of a negative angiogenic factor creating a “no-go” region at the incipient fovea. Viewed together, these findings suggest that a focal concentration of an antiproliferative factor at the incipient fovea defines the avascular region. 

An unexpected finding of this study was the observation that by Fd105 both the stellate astrocytes and blood vessels of the GCP formed a ring that was spatially coincident with the rim of the early foveal depression. Inner retinal neurons were displaced peripherally toward the foveal rim as the foveal depression was formed, but the mechanisms underpinning these displacements remain enigmatic. ^{1 2 4 48} The spatial coincidence of the astrocyte–vascular ring and the foveal rim raised the possibility that whatever factor(s) defined the GCP no-go zone may also act to exclude or repel neurons from within the incipient fovea. If such a factor were activated around the time that the avascular area is defined, it could trigger the commencement of the peripheral displacements forming the foveal depression. Alternatively, we have suggested previously that inner retinal neurons within the avascular area may be metabolically stressed, because they are very numerous but must rely on diffusion of oxygen from the relatively distant choriocapillaris. ⁴ Inner retinal neurons may therefore be triggered to migrate toward the vascular ring to resolve their metabolic needs. 

Our observation that from approximately Fd105 GCP astrocytes became less numerous in central monkey retina is consistent with a previous report on postnatal retina, ⁴⁹ although in the present study, this decline in astrocyte populations began prenatally. The density of astrocytes in postnatal monkey central retina declined from 400/mm² at P day 1 to 23/mm² in the adult, associated with loss of GFAP-IR around the fovea. ⁴⁹ This observation strongly suggests that stellate astrocytes play an important but transient role in establishing the perifoveal plexus. What remains to be determined is whether the disappearance of astrocytes occurs by programmed cell death, withdrawal to the periphery, and/or downregulation of GFAP to undetectable levels. This could resolve whether astrocytes are not necessary to sustain the perifoveal plexus or are still present but undetectable and could respond if needed. 

Implications for Foveal Development in Premature Infants

In the present results, definition of the avascular area, deepening of the foveal depression by neuronal and Müller cell migration, and maturation of the perifoveal plexus took place between Fd105 and P3 week in the Macaca monkey. In the human a comparable period is 25 WG to P3 month. ²⁵ Human infants are at high risk of development of retinopathy of prematurity if delivered before 28 WG. ⁵⁰ Long-term follow-up studies indicate that even when retinopathy is not present in the neonatal period, premature infants of 31WG or less are at increased risk of development of some form of ocular disorder, including myopia, hypermetropia, strabismus, astigmatism, or poor visual acuity. ^{50 51 52 53 54} The results of the present study suggest that the astrocyte–vascular interactions that establish the human perifoveal plexus occur during this vulnerable period, as do the neuronal migrations that form the foveal depression and elevate cone density. ^{4 28} Because perifoveal plexus development and formation of the foveal depression appear interdependent, it seems likely that the changes in retinal Po ₂ experienced by premature infants could affect these normal developmental interactions, retarding growth of the perifoveal plexus at a time when the foveal depression is beginning to form. This in turn may indirectly affect neuronal displacements, resulting in subtle abnormalities that may underlie the visual difficulties experienced by premature infants. 

 Supported by the Ophthalmic Research Institute of Australia (JMP), Sydney Eye Foundation (TS), National Institutes of Health Grant EY04536 (AEH), and Research to Prevent Blindness (AEH).

 Submitted for publication May 27, 1999; revised March 7, 2000; accepted March 17, 2000.

 Commercial relationships policy: N.

 

Figure 1.

 

(A, B) The same Fd88 double-labeled section in which the optic disc is to the left and the incipient fovea is to the right, both beyond the field of view. (A) Astrocytes labeled by GFAP were found between the optic disc and incipient fovea. They were abundant near the optic disc in the nerve fiber layer (NFL) and GCL; but close to the fovea (arrow), astrocytes were found only at the GCL-IPL interface. Very few astrocytes were present in the NFL more than 1 mm from the optic disc (∗). (B) Blood vessels labeled by CD31 were present in the NFL and GCL near the optic disc; but close to the incipient fovea, vessels were present only at the GCL-IPL interface (arrowhead). Astrocytes in (A) were 150 μm closer to the fovea than blood vessels. (C, D) The same double-labeled section through the developing Fd101 fovea showing blood vessels labeled by CD31 in (C) and astrocytes labeled for GFAP in (D). The avascular area (bracket) overlies the foveal cone mosaic (fcm). Its edges are defined by the sparse blood vessels (arrowheads) and astrocytes (arrows) of the GCP. The GCP astrocytes reached slightly farther into the incipient fovea than vessels.

Figure 2.

 

Confocal images from the paired eyes of an Fd105 monkey. Each retina has been double labeled to demonstrate blood vessels detected by CD31 (red) and GFAP staining in glia (green). (A) A frozen section through the developing fovea shows that a shallow depression appeared in the GCL centered within the avascular region. Arrowheads at the top indicate the margin of the depression, and lower arrowheads mark the GCP vessels labeled with CD31. Astrocytes extended slightly farther into the foveal avascular zone on the right. Note the increased labeling for GFAP in the inner Müller cell processes within the fovea. (B) Merged confocal images from a wholemount of the other Fd105 eye showing that astrocytes and blood vessels in the GCP surrounded the avascular developing foveal depression. Many blind-ending capillary sprouts protruded from the capillary bed toward the avascular region. These sprouts were invested with astrocytes (arrowhead) that extended 25 to 50 μm farther into the fovea. Punctate labeling for GFAP within the avascular region was in Müller cell foot processes. The small arrow points out the same blood vessel loop in (B) and (C). (C) A low-magnification single-image confocal view of the GCP plexus and its larger feeder vessels (fine arrows). Thick arrow: Direction of the optic disc in (C, D). (D) The corresponding confocal image showing the overall astrocyte distribution. In the periphery, their distribution did not match the vascular pattern, although there was some investment of larger vessels (thin arrows). Astrocytes are clustered in the GCP (arrowhead).

Figure 3.

 

(A) Double-label immunofluorescence in an Fd105 section, created by overlaying two conventional micrographs and using a computer program to pseudocolor one. Müller cells were heavily reactive for CRALBP (green) and had a predominant vertical orientation. One large, stellate astrocyte contained GFAP (red) and lay at the GCL-IPL interface with a predominant horizontal orientation. (B, C) Immunofluorescent labeling of adjacent Fd140 sections for GFAP in (B) and CRALBP in (C) with retinal layers aligned. Note the marked difference in cell process orientation and the relative low level of GFAP in Müller cells compared with astrocytes. (D through H) Overlaid confocal images of a section double labeled for GFAP (green) and CD31 (red). (D, E) At Fd115 in the region (D) between the fovea and optic disc and (E) near the fovea, branches of the GCP passed through the IPL and were beginning to form the ICP (small arrows) and in (D) a few capillaries were present in the OCP (arrowheads). Astrocytes not associated with capillaries were beyond the GCP. (F) An Fd140 retina near the optic disc had four capillary plexuses. Large stellate astrocytes were common in the GCP, but their processes stayed confined to the inner retina. (G) At Fd132 a very clear gap in the GCP overlay the foveal depression, but the capillary branches into the outer retina were absent. (H) Fd160 foveal edge shows the developing perifoveal capillary plexus in outer retina. Closest to the foveal depression only the ICP was present (arrows), but, more peripherally, capillaries crossed the INL to form the OCP (arrowheads). Vertically oriented Müller cells labeled heavily for GFAP in both (G) and H. Scale bar, (C) 50 μm; (E, H) 100 μm.

Figure 4.

 

Frozen sections through the developing fovea stained by double-label immunofluorescence to show GFAP labeling of glia (left) and CD31 labeling of blood vessels (right). (A, B) Both the inner processes of Müller cells within the avascular area and GCP astrocytes closest to the depression were labeled for GFAP. The arrowhead and arrow indicate the farthest point that astrocytes or blood vessels penetrated the fovea. (C, D) Three weeks before birth the astrocyte-free region (arrowheads) measured approximately 500 μm, and the vessel-free region (arrows) was only 200 μm wide. Vessels in the GCP lay closer to the depression than either the OCP or ICP vessels (thick arrows). (E, F) Very few astrocytes were present centrally after birth (arrowhead). The GCP (arrow) and OCP (thick arrow) were present at the foveal edge, but the ICP still had not formed there. (G, H) After birth the perifoveal capillary plexus was formed by the anastomoses of GCP, ICP, and OCP plexuses and surrounded the fully formed foveal pit at the level of the ICP (arrows). No astrocytes were present in the foveal region, although some lay in the GCP outside the fovea (arrowhead). Müller cells labeled heavily for GFAP in both their inner and outer processes around birth.

Figure 5.

 

Confocal images of retinal flatmounts with blood vessels labeled by a mixture of monoclonal antibodies to CD31 and von Willebrand factor (red). Astrocytes and Müller cells were labeled for GFAP (green). (A, B) Two focal levels of blood vessels surrounding the fovea. The dense GCP plexus (A) surrounded the foveal depression with short, blind-ending vessels extending onto the foveal rim. The developing ICP and OCP (merged in B) were sparse compared with the GCP. They lay some distance from the fovea, which was marked by GFAP labeling (green) in Müller cells. The large arrow indicates the direction of the optic disc in (A) and (B). (C) High-power merged images of the GCP. Many short, blind capillaries still protruded into the avascular area (short arrows), but most astrocytes (thin arrows) lay 100 to 150 μm peripheral to the perifoveal ring. Müller cell end feet were GFAP-IR in the foveal pit. (D) An image of the ICP-OCP was duplicated and pasted into the blue channel of an image analysis software document, giving it a pink-purple appearance. The GCP was red. Both lay close to the fovea, which was slightly distorted by a fold that artifactually labeled red. Anastomoses between the three capillary beds (arrows) were common within 200 to 300 μm of the perifoveal ring. In the juvenile retina the inner (E) and outer (F) capillary beds forming the perifoveal plexus were coincident, and points of anastomoses were common (arrows). GFAP labeling in foveal Müller cells had almost disappeared and only a few stellate astrocytes (green cells in E) were detected in the inner retina. Scale bar, (B, F) 100 μm.

The authors thank the staffs of the Regional Primate Research Center at the University of Washington and the Indonesian Primate Center (Bogor, Java) for their critical role in obtaining this valuable primate tissue; and Andra Erickson, Dan Possin, and Vera Terry for technical assistance.