Vaso-Obliteration in the Canine Model of Oxygen-Induced Retinopathy

Transcription

1 Vaso-Obliteration in the Canine Model of Oxygen-Induced Retinopathy D. Scott McLeod, Rachel Brownstein, and Gerard A. Lutty Purpose. To quantify the acute constrictive response of developing retinal blood vessels to hyperoxia and to examine the vaso-obliterative phase of sustained oxygen breathing in the neonatal dog model of retinopathy of prematurity. Methods. Seven littermates were used to examine the acute constrictive response of the developing retinal vessels to hyperoxia (30 minutes to 96 hours of 100% oxygen). ADPase retinal flatmounts were prepared, and morphometric measurements were made using computerassisted analysis. Vaso-obliteration also was examined in three animals killed after prolonged exposure to hyperoxia (4 days of 100% oxygen) and in three room air controls using ADPase flat-embedded retinas and cross-sections. Choroids were processed for alkaline phosphatase flat-embedding. Results. After 1 hour of oxygen breathing, all vascular components showed a reduction in diameter: Arteries were reduced 27%, veins 18.3%, and capillaries 27.7%. Capillary constriction peaked by 24 hours (69.4% reduction), whereas arteries and veins continued to close. Although capillary diameters did not decrease significandy after 24 hours, the number of capillaries, as determined by percent vascular area calculations, continued to decrease in all areas through die additional 3 days of oxygen breathing. In contrast, after 4 days of hyperoxia, the choriocapillaris lumenal diameters and percent vascular area did not vary significandy from controls. Analysis of sections taken through various retinal regions of these animals revealed significant decreases (40%) in the volume of die extracellular spaces available for blood vessel formation. Hyperoxia also resulted in a 55.6% decrease in the total number of cells (endothelial cells, ablumenal cells, perivascular cells) within the inner retina; however, there was no significant difference in ganglion cell counts in the two groups. Conclusions. This study demonstrates diat the pattern and severity of the reaction of developing retinal vessels to hyperoxia in the newborn dog is similar to that described for the kitten and the premature human. This response is unlike that exhibited by the newborn rat or mouse. Invest Ophthalmol Vis Sci. 1996;37: JDuring normal retinal vasculogenesis, primary vascular remodeling occurs in the form of capillary constriction, retraction, and atrophy, which progressively eliminates redundant channels. It has been suggested that the response of the immature retinal vessels to high oxygen tension, as occurs in retinopathy of prematurity (ROP), is an exaggeration of this process that results in wholesale capillary closure. 1 Ashton et al 1 From the The Wilmer Ophlhnlmological Institute, Johns Hopkins University School of Medicine, Baltimore, Maryland. Supported by National Institutes of Health grants EY09357 (GAL) and EY01765 and by Research to Prevent Blindness. GAL is an American Heart Association Established Investigator. Submitted for publication June 7, 1995; revised August 29, 1995; accepted October 4, Proprietary interest category: N. Reprint requests: Gerard A. Lutty, Johns Hopkins Hospital, 170 Woods Research Buildi7ig, 600 N. Wolfe Street, Baltimore, MI) demonstrated that in kittens exposed to hyperoxia, there was a gradual withdrawal of endothelial cells from the primitive capillaries toward parent vessels, which at more advanced stages demonstrated degenerative changes. He suggested that oxygen-induced vaso-obliteration could be compared to normal vascular remodeling, and he considered that the ultimate pattern of normal capillary growth may in some way be related to oxygen tissue tension. 2 Although a constrictive response to high oxygen concentrations also is exhibited by mature adult retinal vessels, it is the immature retinal vasculature that is susceptible to oxygen-induced obliteration. In contrast, it has been suggested that the more mature neonatal choroidal circulation remains intact, with essentially normal blood flow. 3 Ashton et al 1 and Patz 4 suggested that relative oxy- 300 Investigative Ophthalmology & Visual Science, February 1996, Vol. 37, No. 2 Copyright Association for Research in Vision and Ophthalmology

2 Oxygen-Induced Vaso-Obliteration 301 gen deficiency is probably one of the normal stimuli for extension of retinal vessel growth anteriorly. Additionally, Patz noted, "it is apparent that during continuous oxygen administration, the oxygen tension in the retina is increased by diffusion from the choroid. Thus, during oxygen administration, this stimulus is removed, and retinal vessel growth is thus suppressed with ultimate obliteration of retinal vessels." 4 The proliferative phase of ROP occurs after withdrawal from a high-oxygen environment, when it is thought the inner retina, previously hyperoxygenated, becomes hypoxic as a result of the decreased diffusion of oxygen from the choroidal blood and the loss of retinal vessels. 1 ' 4 This results in an abnormal overgrowth of blood vessels in the inner retina. Ashton et al 1 suggested that ROP was probably no more than a violent activation of the normal process of retinal vascularization, so that the narrow confines of the nerve fiber layer could no longer contain the exuberant vasoformative tissue.' Ashton et al 1 were the first to demonstrate in an animal model that obliteration of retinal vessels occurred in prolonged hyperoxia and that the response was directly proportional to the degree of immaturity of the vasculature, the duration of exposure, and the degree of oxygen concentration. Studies have demonstrated that the developing retinal vessels of different species react differently to hyperoxia and that variations in this response may occur in the same species, depending on the relative maturity of its retinal vasculature. 5 Although oxygen-induced vaso-obliteration has been studied in the newborn kitten, rabbit, rat, and mouse using a variety of techniques, 5 " 8 this phenomenon has not been investigated fully in the dog model of oxygen-induced retinopathy. In this current study, we examined the morphometric and morphologic features of the initial constrictive response and the process of vaso-obliteration in the neonatal dog model of oxygen-induced retinopathy using computer-assisted analysis of ADPase flatmounted and flatembedded retinas. Vascular morphology in the flat perspective subsequently was correlated with blood vessel structure in histologic sections from flat-embedded tissue. Additionally, the choriocapillaris was analyzed in alkaline phosphatase preparations to determine the effect of hyperoxia on the morphology and structure of the neonatal choroidal vasculature. MATERIALS AND METHODS Vasoconstriction Seven neonatal, purebred, 4-day-old beagles, (345 to 404 g body weight) were used in this portion of the study. One animal served as an air control, and the others were exposed to 100% humidified oxygen for periods ranging from 30 minutes to 96 hours. All procedures were performed in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Animals were killed by an intraperitoneal injection of sodium pentobarbital, and retinas were prepared as ADPase flatmounts. 9 Blood vessels were visualized in specimens using a Zeiss (Oberkochen, Germany) photomicroscope with brightfield illumination. For morphometric measurements, a video signal was obtained from a black-and-white charge coupled device camera (C ; Hamamatsu, Hamamatsu City, Japan) mounted on the microscope, and images were digitized using a Data Translation (Marlboro, MA) image analysis board in a Macintosh Ilci personal computer (Apple, Cupertino, CA). Analysis was performed using National Institutes of Health Image software (version 1.47). Artery and vein measurements were made on images of flatmounts at a location just before the first major bifurcation on the superior arcade (1.7 to 2.3 mm from the optic nerve head) in all specimens. Capillary measurements were made in the central vascularized region of the superior arcade approximately 3 mm from the optic nerve head. At least 20 measurements of capillary diameters for three adjacent fields were made for each specimen. The field of capillaries chosen for analysis was an immature capillary plexus. Percent vascular area determinations were made on the same fields as those used for capillary diameters. These determinations were representative of the capillary density in a given area of retina as determined by initial analysis of random samples of retinal areas. Measurements of percent vascular area were made from binary images of fields of immature capillaries in similar regions of retina from each animal (Fig. IB). Grayscale images (original data set) were converted to binary using the thresholding function of the Image program. The original data set, as well as photomicrographs, were used as a guide for setting threshold before conversion of the image to binary (vessels as black pixels and background as white pixels). The number of black and white pixels were measured using the program to provide a ratio for calculating percent vascular area. Percent vascular area was determined by dividing the number of black pixels (representing blood vessels in the bright-field images) by the total number of pixels in an image. In all cases, identical regions from the two groups of animals were compared. Vaso-obliteration In another experiment, we examined the morphometric and morphologic features of vaso-obliteration in the neonatal dog using ADPase flat-embedded retinas. Six animals (5 days of age at death) were used: Three were killed in hyperoxia (4 days of 95% to 100% oxygen), and three served as room air controls. Body weights at death were 372 ± 77 g for the air control

3 Investigative Ophthalmology & Visual Science, February 1996, Vol. 37, No. 2 lhr 2Hrs 4 Hrs 6 Hrs 24 Hrs 96 Hrs Time in Oxygen FIGURE l. Morphometric analysis (A,B) ar d morphologic (C-F) appearance of the constrictive response of developing retinal vessels to hyperoxia. (A) Plots of mean capillary diameters and SEM in an air control and in animals killed in oxygen after various lengths of time. Data were derived from 20 measurements of capillary diameters from three adjacent fields in the same region of retina from each animal, and values are expressed in //m ± SEM. (B) Percent vascular area plots of data derived from binary images of identical regions in retinas from an air control and animals sacrificed in oxygen (mean ± SEM, derived from three measurements in each animal). (C) Appearance of the neonatal retinal vasculature in the air control animal. (D) Similar area in an animal after 6 hours of oxygen breathing demonstrates constriction of major vessels with obliteration and retraction of some capillaries. (E) Advanced constriction with progressive obliteration and capillary retraction in an animal after 24 hours of oxygen breathing. (F) Extreme constriction and obliteration, resulting in the formation of capillary islands in an animal exposed to oxygen for 4 days. (C-F, curved arrow vein, straight arrow artery; ADPase flatmounts) Magnification, X53. group and 278 ± 59 g for the oxygen-treated group. Animals were killed by intraperitoneal injections of sodium pentobarbital In addition to retinal vasculature, we examined the choroidal vasculature of animals using the alkaline phosphatase flat-embedding technique. 10 Initial analysis was performed in the flat perspective, and then preparations were serially sectioned. For morphometric analysis of retinas, blocks of ADPase flat-embedded specimens were illuminated with dark-field illumination on a Zeiss Photomicroscope II. We analyzed three separate regions from each lobe of each animal (three fields per region). Photomicrographs were made of entire blocks for mapping the subsequent sectioning process, and digitized images were collected for morphometric analysis using the image analysis system and methods described in the previous section. All morphometric analysis of these retinas was performed en bloc. Sections were developed in ammonium sulfide and counterstained with thionin or were stained with periodic

4 Oxygen-Induced Vaso-Obliteration 303 FIGURE 2. Photographic montage of the temporal lobe in an air control animal (A) and an animal after 4 days of exposure to hyperoxia (B). AD- Pase flat-embedded retinas. Magnification, X14. acid-schiff and hematoxylin or with hematoxylin and eosin. Tissue sections from identical regions of control and oxygen-treated retinas were compared. Cell counts were performed manually. Morphometric analysis of the choriocapillaris was performed, before embedding, on wet flatmount specimens using bright-field illumination and the image analysis system described above. We analyzed capillary diameters in the posterior pole, equator, and peripheral choroid, as well as percent vascular area for these regions. In the case of capillary diameters, we measured 20 capillaries in three randomly selected fields for each region. Percent vascular area determinations also were made from three randomly selected fields for each region by converting digital grayscale images to binary images as described in the previous section. RESULTS Vasoconstrictive Response Morphometry. After 1 hour of oxygen breathing, all vascular components showed a reduction in diameter, with arteries reduced by an average of 27.04% less than air controls, veins by 18.3%, and capillaries by 27.7%. By 4 hours, all components of the vascular hierarchy displayed a similar percentage reduction in diameter, with arteries reduced by an average of 40.6%, veins by 43.9%, and capillaries by 48.6% (Fig. 1 A). At 24 hours, the arteries showed the least reduction (7.5%) from the 4 hour animal (48.1% reduction compared to 40.6%), whereas vein and capillary diameters became even more severely reduced veins by an additional 20.3% and capillaries by an additional 20.9%. By 96 hours, arteries were reduced in diameter from the 24-hour animal by an additional 11.4%, veins were reduced an additional 8.2%, and capillaries were reduced only an additional 0.6%. Therefore, in this experiment, artery diameters were reduced from 35.6 to 14.4 /xm, veins from 97.7 to 27.9 /im, and capillaries from 15.6 to 4.7 ^tm. Although capillary diameters were not reduced significantly by the additional 3 days of hyperoxia (Fig. 1 A), the number of capillaries that remained was significantly less. Figure IB shows the percent vascular area determinations for similar fields of capillaries from an air control and from an oxygen-exposed animal at 6 hours, 24 hours, and 96 hours. We calculated the following percent vascular areas in similar regions of retina for these animals to be: air = 62% vascularized, 6 hours = 40.4% vascularized, 24 hours = 21.2% vascularized, and 96 hours = 11.8% vascularized. Note that although capillary diameters (Fig. 1A) were similar in the 24- and 96-hour animals, the percent vascular area and, therefore, the number of remaining viable capillaries (ADPase active) in the 96-hour animal retina had been reduced significantly by an additional 44% from the 24-hour animal. Pattern of Vasoconstrictive Response. Early vaso-obliteration was observed after 6 hours of oxygen breathing (Fig. ID). The obliterative response was restricted initially to the most immature peripheral vasculature and involved closure of some capillary channels and retraction of cells (lumenal and ablumenal) toward other

5 304 Investigative Ophthalmology & Visual Science, February 1996, Vol. 37, No. 2 Peri papillary Central Periphery Peripapillary Central Periphery capillary segments. This ultimately resulted in the formation of capillary islands, viable capillary segments segregated from intact vasculature. These structures always contained trapped red and white blood cells. By 24 hours (Fig. IE), vaso-obliteration progressed from the most immature capillaries of the peripheral retina to more mature capillaries adjacent to the major vessels and within the peripapillary region. Therefore, the constrictive response in the dog was generally uniform, affecting all lobes, regions, and components of the developing vascular hierarchy. Morphometry of Endpoint Vaso-obliteration Area, Extent, and Pattern of Vascularization. The normal neonatal dog vasculature has a basic three-lobed pattern, with several artery and vein pairs emanating from the optic nerve head and extending radially toward the superior, nasal, and temporal aspects of the periphery. As vasculogenesis proceeds, the lobes become less distinct as they expand in area and merge. In the normal 5 day-old air control animals, the average area of vascularized retina was 127 mm 2 compared to 51 mm 2 in animals after 4 days of continuous hyper-

6 Oxygen-Induced Vaso-Obliteration 305 FIGURE 3. Morphometric analysis (A,B) and morphologic appearance of a 5-day-old air control animal (C,E,G) and endpoint vaso-obliteration (D,F,H) in a 5 day-old oxygen-treated (4 days of 100%) animal. (A) Capillary diameter measurements (mean ± SEM) from three regions of the temporal retinal arcade (60 measurements per region) from three air control and three oxygen-treated animals. Diameters varied by area in air control animals and were constricted uniformly in the oxygen-exposed animals. (B) Capillary densities (percent vascular area) in three regions of air control and oxygen-treated animals (mean ± SEM). There was no significant regional difference in air control or oxygen-exposed animals, but oxygen-exposed animals were approximately 20% as vascularized as control animals. (C,D) Comparison of the peripapillary region in an air control (C) and in an oxygen-exposed animal (D) demonstrates severe constriction of major arteries and veins, with almost total obliteration of capillaries after sustained oxygen breathing (D). (E,F) Comparison of the central vascularized region of retina in the control (E) and the hyperoxic animal (F). Note the sinusoidal appearance of the developing capillaries associated with this artery and developing vein. Hyperoxia has resulted in total obliteration of the vein, severe constriction of the artery, and obliteration of most of the capillaries. (G,H) Appearance of forming capillaries in the far periphery of the air-reared animal (G) and obliterated remains of peripheral capillaries in the oxygen-treated animal. Note the remnants of vessels in the form of capillary islands in the oxygen-exposed animal. (C H) ADPase flat-embedded retinas. Magnification, X60. oxia (60% less in the hyperoxic animals). The greatest difference in extent of vascularization in the two groups (as measured from the optic nerve head radially to the peripheral aspect of the intact vasculature) was found in the temporal lobe (temporal air control = 7.0 mm; temporal oxygen = 4.5 mm; nasal air = 7.8 mm; nasal oxygen = 6.2 mm; superior air = 7.6 mm; superior oxygen = 6.6 mm). Therefore, in the oxygen group, the extent of vascularization averaged 36% less in the temporal lobe, 16% less in the nasal lobe, and 14% less in the superior lobe. Total retinal area was similar for both groups (data not shown). By 96 hours, much of the capillary bed had been obliterated or had retracted into segregated islands of cells (Figs. IF, 2). Veins were affected more significandy than arteries and, in some cases, remained viable only in the peripapillary region. In regions in which partial obliteration occurred and veins remained intact, arteriovenous shunts were formed. Retinal Capillary Diameters. Capillary diameters in the air control group were 15.7 ± 1.0 fxm in the periphery, 19.6 ±1.0 fim in the central vascularized region, and 11.7 ± 0.5 //m in the peripapillary region, as determined in the flat perspective of ADPase flatembedded specimens. After 4 days of continuous oxygen breathing, diameters of remaining viable capillaries (ADPase active) were reduced compared to airreared animals (Fig. 3) by an average of 62% in the periphery, 61% in the central vascularized region, and 43% in the peripapillary region (Fig. 3A). The vasoobliteration was equivalent in each arcade (temporal, nasal, and superior), and capillary diameter reduction in hyperoxic animals was similar. Therefore, there was uniformity of response to hyperoxia within all regions and lobes of the neonatal dog retina. Determinations of percent vascular area in hyperoxic animals demonstrated that capillary density was reduced compared to air controls by 77% in the peripheral vascularized region, 72.5% in the central vascularized region, and 81.5% in the peripapillary region (Fig. 3B). Choriocapillaries Diameters. The choriocapillaris of the dog has been studied previously." In the dog, as in the human, the choriocapillaris is an extensive anastomosing network of fenestrated capillaries external to the retinal pigment epithelium. In tapetal regions, Bruch's membrane is attenuated extremely, and the capillaries indent the retinal pigment epithelium, possibly as a result of the rigidity of the tapetal cells. In the normal 5-day-old dog, we found that the choroidal vasculature is fully formed. Morphometric analysis of choriocapillaris in different regions of alkaline phosphatase incubated choroids from an air control and from hyperoxia-exposed animals revealed no significant differences in capillary pattern, diameter (Fig. 4A), or density (Fig. 4B, percent vascular area), although the retinas from these animals were affected severely. Additionally, cross-sectional analysis (Figs. 4G, 4H) demonstrated remarkable similarities in structure and cellular components of the choroidal blood vessels. Structural Analysis of Vaso-obliteration Effect of Oxygen on Basic Retinal Architecture. In the 5-day-old air control, the retina consisted of four basic layers: a relatively undifferentiated neuroblastic layer, an inner plexiform layer, a ganglion cell layer, and an innermost nerve fiber layer consisting of spaces formed by the inner Miiller cell processes. Total retinal thickness varied from 215 //m at the nerve head to 165 fim in the far periphery. At this age, the forming

7 Investigative Ophthalmology & Visual Science, February 1996, Vol. 37, No Z > 40- S 20- ft, n. I n Air D Hyperoxia JU I LL Pcripapillary Equator Periphery g Peripapillary Equator Periphery FIGURE 4. Morphometric analysis (A,B) and morphologic appearance (C-H) of the choriocapillaris in three regions of choroid from a 5 dayold room air control animal (CJE.G), and a 5-day-old animal exposed to 100% oxygen for 4 days (D,F,H). (A) Comparison of mean choriocapillaris diameters ± SEM (fim) showing no significant differences in the two animals. (B) Comparison of choriocapillaris density (percent vascular area) in the two animals. There was no significant difference in percent vascular area in control and oxygenexposed animals. (C-F) Morphologic appearance of choriocapillaris in the posterior pole (C,D) and periphery (D,F) from a room air control animal (C,E) and an oxygenexposed animal (D,F). (C-F) Alkaline phosphatase. Magnification, X128. Representative sections from the equatorial tapetal choroid from an air control (G) and an oxygenreared animal (H) demonstrate no appreciable differences in choriocapillaris structure (arrows). Thionin; magnification, X710. outer plexiform layer was just discernible (2 to 3 /zm in thickness), 2 to 3.4 mm from the nerve head, depending on the lobe (the temporal lobe was less developed). Rudimentary inner segments were evident and varied in length from 3.7 (j,m at nerve head to 1.8 /zm in the far periphery. In the hyperoxic animal, we found no differences in the length of inner segments nor in thickness and extent of the developing outer plexiform layer, but we did observe an overall 11% decrease in the total retinal thickness compared to the air control group. This decrease was not related to changes in the outer retinal layers but was associated with a decrease in thickness of the innermost retina. Therefore, hyperoxia did not retard development of photoreceptor inner segments or synaptogenesis of photoreceptors and developing first-order retinal neurons. Muller Cells and Cell-free Spaces of Avascular Retina. Analysis of sections taken through peripheral retina revealed significant decreases in the span (or height) of Muller cell processes from inner nuclear layer to internal limiting membrane after sustained oxygen breathing (Fig. 5), which accounts for the decrease in inner retina mentioned previously. Measurements made from internal limiting membrane to inner plexiform layer demonstrated as much as a 40% reduction in span (or height) of inner retinal Muller cell processes in hyperoxic animals (Fig. 5A). Such was the

8 Oxygen-Induced Vaso-Obliteration 40Y 30 Air D Hyperoxia Avascular Edge of Vasculature Vascular FIGURE 5. Morphometric analysis (A) and morphologic appearance (B,C) of Miiller cell processes and extracellular spaces in the inner retina of a 5-day-old room air control animal (B) and a 5-day-old animal exposed to 100% oxygen for 4 days (C). (A) Comparison of measurements made of the span of inner retinal Miiller cell processes in three regions (mean ± SEM in /im). (B) Sections collected through avascular retina from an air control (B) and an oxygenexposed animal (C) demonstrate decreased height or span of inner Miiller cell processes (arroivs) and reduced volume of extracellular space in the oxygen-exoposed animal. Undifferentiated angioblasts are shown (a). Thionin; magnification, X880. case in avascular retina and in regions of peripheral vasculature that had undergone vaso-obliteration. Morphologically, these processes appeared swollen and more broad in the hyperoxic environment. This change in Miiller cell morphology resulted in a significant decrease in the volume of the extracellular spaces (Figs. 5B, 5C). Capillaries. Normal, immature retinal capillaries in the neonatal dog were composed of a plethora of endothelial cells, with as many as seven endothelial cell nuclei occupying one 2.5-/xm thick cross-section of a capillary lumen (Figs. 6, 7). These cells were rounded and, perhaps because of their sheer numbers, were forced to protrude into the lumenal space. There was a thin, periodic acid-schiff-positive basement membrane, and there were widely scattered, irregularly distributed pericytes. It was obvious from an examination of serial sections that the number of endothelial cells far outnumbered the pericyte population (Fig. 6C), with ratios of up to 7:1 existing in some capillary segments. Blood elements consisting of red blood cells, platelets, and leukocytes were observed frequendy within the immature capillaries (Fig. 7C). Also, cells of unknown phenotype exhibiting various morphologic phases of degeneration were noted within the lumenal space of many capillaries. These degenerative cells displayed a typical apoptotic morphology, i.e., fragmented nuclear material. In hyperoxic animals (Fig. 6D), we found that obliterated capillaries (which had lost ADPase activity) consisted of delicate remnants of basement membrane material (periodic acid-schiff positive), whereas capillary segments that had become severely constricted and yet remained viable (ADPase activity remaining) demonstrated a significant reduction in endothelial cell component. Of note were segments of capillary remnants completely devoid of endodielium yet containing what appeared to be intact pericytes (Fig. 7B). We also examined the structure of capillary islands that formed in the hyperoxic animals: Tangential sections through these structures revealed cellular debris within the lumen of these segregated capillary segments that consisted of degenerating endothelial cells (Fig. 7D). The source of die debris within the structures was unclear, but it might have been the cellular remains of trapped leukocytes, sloughed endothelial cells, or both. Vasoformative and Neuronal Cell Loss. Cross-sections of retinas clearly demonstrated cell loss in inner hyperoxic retina. To analyze the magnitude of this loss (Fig. 6), we counted cell nuclei in the inner retina (ganglion and nerve fiber layers) and in the inner portion of the neuroblastic layer in the vascularized peripheral retina of air controls and in vaso-obliterated regions of peripheral retina in animals killed in oxygen (Figs. 6A, 6B). We counted all cell nuclei in the inner retina

9 308 Investigative Ophthalmology 8c Visual Science, February 1996, Vol. 37, No. 2 FIGURE 6. Morphometric analysis (A,B), and morphologic appearance (C,D) of cell loss in the hyperoxic retina. (A) Comparison of total cell counts and ganglion cell counts in peripheral vascularized region of inner retina from a 5-day-old air control animal and vaso-oblite rated peripheral region from a 5-day-old animal exposed for 4 days to 100% oxygen (mean cell counts ± SEM). Ganglion cell counts for these same regions were: air = 2.7 ± 1.0 cell nuclei per high-power field; oxygen = 2.4 ± 0.8 cell nuclei per high-power field, (B) Comparison of inner neuroblastic layer cell counts in a 5-day-old air control animal and a 5 day-old oxygen-exposed animal (4 days, 100% oxygen). Photomicrographs of sections taken through the peripheral vascularized inner retina of a 5-day-old air control animal (C) and a peripheral vaso-obliterated region from a 5-day-old oxygen-exposed animal (D). (C) In the air control animal, numerous endothelial cells occupy the immature capillary lumen {arrows). The inner neuroblastic layer is shown (curved arrow). (D) In this field from a hyperoxic animal, no endothelial cells remain in obliterated capillaries, which consist primarily of a thin remnant of basement membrane material (long arrows). A ganglion cell is present (short arrow), and the inner neuroblastic layer, with its more loosely arranged appearance, is shown (curved arrow), note the narrowing of the inner retina. (C,D) Periodic acid-schiff (PAS) and hematoxylin; magnification, X490. (lumenal endothelial cells, ablumenal cells, perivascular cells, and ganglion cells) in four randomly selected high-power fields. Twenty fields per section from 25 separate sections were counted for each animal; sections were separated by a distance of 12.5 fj,m. In each case, we noted the number of definitive ganglion cell bodies located in each field. Analysis revealed that compared to air rearing, sustained oxygen breathing resulted in a 55.6% decrease in the total number of cells within the inner retina (Fig. 6A). Comparison of ganglion cell counts, however, revealed no significant differences (Fig. 6A) in the two groups. Additionally, we observed a looser arrangement of cells in the innermost portion of the neuroblastic layer of hyperoxic animals. Analysis of the cell density within these regions (Fig. 6B) revealed a 27.7% decrease in the number of cells in the hyperoxic neuroblastic layer. These results demonstrate that cell death occurs in the retina of the oxygen-breathing animal (Fig. 6D) and that cells involved in vasculogenesis show more significant vulnerability to high oxygen than neuroblasts. DISCUSSION Comparison of Oxygen Response to Other Models This study has demonstrated that the pattern and severity of the reaction of developing retinal vessels to hyperoxia in the newborn dog is similar to that described for the kitten and the premature human. 512 This pattern of response is unlike that exhibited by the newborn rat 1314 or mouse (5 days of age at time of hyperoxic insult), 8 in which vaso-obliteration occurs primarily around the disc and gradually diminishes toward the periphery. 5 Although few studies have quantified thoroughly the oxygen response in the human or animal models of retinopathy of prematurity,

10 Oxygen-Induced Vaso-Obliteration 309 FIGURE 7. (A) Section demonstrating the numerous endothelial cells {straight arrows) associated with immature capillaries in the normal 5-day-old animal. A pericyte nucleus is shown {curved arrow). Periodic acid-schiff (PAS) and hematoxylin; magnification, X1050. (B) Obliterated capillary in the oxygen-treated animal demonstrates complete loss of endothelium, with an intact pericyte {curved arrow). PAS and hematoxylin; magnification, X1050. (C) Tangential section showing the appearance of healthy endothelial cells {solid arrows) in a normal air control animal. Leukocytes are also shown {open arrows). PAS and hematoxylin; magnification, X760. (D) Tangential section through a segregated capillary island in the hyperoxic animal contains degenerating endothelial cells {short arrows) and cellular debris {long arrow). PAS and hematoxylin; magnification, X570. Penn and Gay 6 did demonstrate that the major retinal vessels of the newborn rat react with modest constriction, even with extended periods of high oxygen breathing (14 days of exposure to continuous 80% oxygen). In the rat, they found an 8% reduction in artery diameter and a 22% reduction in vein diameter in india ink-perfused flatmounts, compared to 60% and 72% reductions, respectively, in the dog. Although the oxygen concentration in our study was 20% higher than in the aforementioned study, the duration of exposure was significantly less (72% less exposure time). Unfortunately, the rat study did not quantify the constrictive response at the capillary level; therefore, results from this aspect of our study cannot be compared with the rat. Although we recognize that ADPase activity is localized on the ablumenal surface of the endothelial cells in these preparations and may not be an accurate measure of lumenal diameters, we compared morphometric results obtained with AD- Pase processed and india ink-infused retinas (the tracer method used in the rat study, which only reveals the lumenal space) from normal 5-day-old air control dogs (unpublished data, 1987). Although capillary diameters were slightly smaller in dogs perfused with ink, statistically the results were not significantly different (ADPase average = 16.6 ± 3.5 jum, india ink = 14.1 ± 1.2 fxm). Muller Cell Processes and Extracellular Spaces Extracellular spaces in avascular peripheral retina have been described in the premature human In the avascular neonatal dog retina, the Miiller cell processes and the spaces they provide are important structures for support during angioblast migration, blood vessel assemblage, and unencumbered growth anteriorly Additionally, early vascular form and patterning may be dictated somewhat by the arrangement of Miiller cell processes and spaces they create during the period of vasculogenesis. 17 In hyperoxic animals, we found a significant reduction in the span of these processes in the inner retina. In fact, these spaces become obliterated after return to normoxia. 19 This observation is important because before the revascularization process that accompanies return to normoxia, the volume of extracellular space available for assemblage of blood vessels and unencumbered growth anteriorly has been reduced significantly. Of particular significance is the observation that these

11 310 Investigative Ophthalmology & Visual Science, February 1996, Vol. 37, No. 2 spaces are never recovered in oxygen-treated animals, 19 an observation that has been made in premature humans 16 with Vasoformative Cell Loss In this study, we observed an apparent differential effect of hyperoxia on endothelial cells and pericytes of immature capillaries. In normal development, endothelial cells dominate the cell population of capillaries: Ratios as high as 7:1 of endothelial cells to pericytes may exist in some capillary segments. In the adult retina, this ratio is 1:1. 20 Therefore, during retinal vasculogenesis, redundancy of endothelial cells exists and is probably modified during the process of normal vascular remodeling, which takes place with continued development. In hyperoxic animals, we found a significant loss of cells in the inner retina in which vasoobliteration occurred (56% reduction), suggesting that during the revascularization process that accompanies return to room air, proliferation will be required to replenish the pool of degenerated vasoformative cells. In support of this, recent studies 19 from our laboratory indicate that vasoproliferative activity is extremely low during normal vasculogenesis, supporting the view that retinal blood vessels form by a process of in situ differentiation and organization of angioblasts. 17 During the revascularization that occurs after hyperoxic insult, however, vaso-proliferative activity is extremely high. 19 Of interest was the observation of intact pericytes associated with obliterated segments of immature capillaries in the hyperoxic animals. In vitro studies have demonstrated that endothelial cells are more sensitive to high oxygen than are pericytes. 21 ' 22 Additionally, in vitro assays said to mimic differentiation have shown that endothelial cells at confluence or after assemblage into tube-like formations demonstrate increased superoxide dismutase activity and, therefore, are more resistant to oxygen-induced damage. 23 Capillary Islands and Arteriovenous Shunts Capillary islands, like those in the hyperoxic retina of the neonatal dog, have been observed in the retina of oxygen-exposed kittens 7 ' 12 and in premature humans administered supplemental oxygen. 12 Perhaps as a result of trapping blood cells in the process of closure, endothelial cell metabolism in these structures can somehow be maintained. In regions of dog retinas demonstrating partial capillary obliteration, the circulation was maintained by formation of arteriovenous shunts. These shunts, considered to be preferential channels for blood flow through the capillary bed, have been observed in the kitten and the premature human as well and are the last to close in hyperoxia. 12 Choriocapillaris Fluorescein angiographic studies have shown that although cessation of blood flow occurs in the immature retina of the kitten during sustained oxygen breathing, blood flow through the choroid appears unaffected. 3 This supports the view that the high oxygen concentration of blood diffusing from the unaffected choriocapillaris in the oxygen-breathing animal (approximately 450 mm Hg PaO 2 in the newborn dog) may satisfy the relatively low metabolic requirements of the undifferentiated neural retina (undifferentiated photoreceptors and neurons). Indeed, in the neonatal dog retina, inner retinal neurons have not yet differentiated fully (as determined by morphologic criteria), and synaptogenesis with developing photoreceptors has only begun (in the peripapillary region). Our analysis demonstrates that in the newborn dog, based on morphologic criteria, the choroidal vasculature is complete and appears unaffected by hyperoxia. Therefore, in the dog, it seems fair to assume that retinal vaso-obliteration is a manifestation of an abnormally high supply and a relatively low demand for oxygen during the developmental period. Perhaps this was best demonstrated in the kitten model by Ashton et al, 1 in which vaso-obliteration did not occur during sustained oxygen breathing where the retina was detached from the choroid and retinal vessels supplied all oxygen. Our study presents the first experimental data to demonstrate morphologic similarities of the choroidal microcirculation in the normoxic and hyperoxic environments. In summary, our results demonstrate that the pattern and severity of the response of the newborn dog to hyperoxia is similar to what has been reported for the premature human. This response is characterized by an overall 60% reduction in diameter of surviving capillaries, a 77% decrease in capillary density, and a 56% loss in vasoformative elements. Although the developing retinal vessels are extremely vulnerable to oxygen-induced damage, the more mature choroidal vessels in the dog appear unaffected, supporting the view that the diffusion of oxygen from the choroid is sufficient to satisfy the relatively low metabolic demands of the immature retina. The newborn dog and the premature human share many of the same morphologic features of vaso-obliteration; therefore, this model is well suited to experimental studies of retinopathy of prematurity. Key Words choriocapillaris, retinal development, retinal vasculature, retinopathy of prematurity References 1. Ashton N, Ward B, Serpell G. Effect of oxygen on developing retinal vessels with particular reference to

12 Oxygen-Induced Vaso-Obliteration 311 the problem of retrolentalfibroplasia.brjophthalmol. 1954; 38: Ashton N, Pedler C. Studies on developing retinal vessels: IX. reaction of endothelial cells to oxygen. Br J Ophthalmol. 1962;46: Patz A. The role of oxygen in retrolental fibroplasia. Trans Am Ophthal Soc. 1968; 66: Patz A. Oxygen studies in retrolental fibroplasia: IV: Clinical and experimental observations. Am] Ophthalmol. 1954; 38: Ashton N. Some aspects of the comparative padiology of oxygen toxicity in the retina. Br J Ophthalmol. 1968; 52: Penn JS, Gay CA. Computerized digital image analysis of retinal vessel density: Application to normoxic and hyperoxic rearing of the newborn rat. Exp Eye Res. 1992;54: Chan-Ling T, Tout S, Hollander H, Stone J. Vascular changes and their mechanisms in the feline model of retinopathy of prematurity. Invest Ophthalmol Vis Sri 1992; 33: Smith LEH, Wesolowski E, McLellan A, et al. Oxygeninduced retinopathy in the mouse. Invest Ophthalmol Vis Sri. 1994;35: Lutty GA, McLeod DS. A new technique for visualization of the human retinal vasculature. Arch Ophthalmol. 1992; 110: McLeod DS, Lutty GA. High resolution histologic analysis of the human choroidal vasculature. Invest Ophthalmol Vis Sri. 1994; 35: Lesiuk TP, Braekevelt CR. Fine structure of the canine tapetum lucidum. J Anat. 1983; 136: Ashton N. Oxygen and the growth and development of retinal vessels. Am] Ophthalmol. 1966; 62: Penn JS, Tolman BL, Henry MM. Oxygen-induced retinopathy in the rat: relationship of retinal nonperfusion to subsequent neovascularization. Invest Ophthalmol Vis Sri. 1994;35: Reynaud X, Dorey CK. Extraretinal neovascularization induced by hypoxic episodes in the neonatal rat. Invest Ophthalmol Vis Sri. 1994;35: Foos RY. Retinopathy of prematurity: Pathologic correlation of clinical stages. Retina. 1987; 7: Kushner BJ, Essner D, Cohen IJ, Flynn JT. Retrolental fibroplasia II. Pathologic correlation. Arch Ophthalmol. 1977;95: McLeod DS, Lutty GA, Wajer SD, Flower RW. Visualization of a developing vasculature. Microvasc Res 1987; 33: Flower RW, McLeod DS, Lutty GA, Goldberg B, Wajer SD. Postnatal retinal vascular development of the puppy. Invest Ophthalmol Vis Sri. 1985; 26: Crone S, Lutty GA, Merges C, McLeod DS. Morphometric analysis of proliferation in the normal developing primary retinal vasculature and in revascularization following hyperoxia. ARVO Abstracts. Invest Ophthalmol Vis Sri. 1995;36:S Hogan MJ, Alvarado JA, Weddell JE. Histology of the Human Eye. Philadelphia: WB Saunders; 1971: D'Amore PA, Sweet E. Effects of hyperoxia on microvascular cells in vitro. In Vitro Cell Dev Biol. 1987;23: Rosen P, Boulton M, Moriarty P, Khaliq A, McLeod D. Effect of varying oxygen concentrations on the proliferation of retinal microvascular cells in vitro. Exp Eye Res. 1991; 53: Smith LEH, Sweet E, Freedman S, D'Amore PA. Alterations in endothelial superoxide dismutase levels as a function of growdi state in vitro. Invest Ophthalmol Vis Sri. 1992;33:36-41.