5′ Nucleotidase and Adenosine during Retinal Vasculogenesis and Oxygen-Induced Retinopathy

Abstract

purpose. 5′ nucleotidase (5′N) is a major source of the vasogenic substance adenosine in most tissues. The distribution and relative levels of 5′N and adenosine were examined in neonatal dog inner retina during normal vasculogenesis and oxygen-induced retinopathy (OIR).

methods. Animals ranging in age from 1 to 22 days of age were used in this study. Adenosine immunolocalization was performed on frozen sections with an antibody against adenosine conjugated to laevulinic acid using a streptavidin peroxidase technique. Triplicate room air control animals at different postnatal ages and triplicate oxygen-treated animals at different time points during or after hyperoxic insult were analyzed. Adenosine immunoreactivity (AI) and 5′N enzyme histochemical reaction product were quantified using microdensitometry. Adjacent sections were incubated for von Willebrand factor to label blood vessels.

results. During normal vasculogenesis, AI was most prominent within the inner retina. The peak of immunoreactivity was located at the border of vascularized retina throughout the period of primary retinal vasculogenesis (1–15 days of age). At 22 days when vasculogenesis was complete, AI decreased within the inner retina. The highest 5′N activity was localized to inner Muller cell processes in inner retina and decreased after vasculogenesis was complete. In animals killed after 4 days of oxygen breathing, the vaso-obliterative stage of OIR, AI and 5′N activity were reduced throughout the retina. During the vasoproliferative stage, AI was markedly elevated at the edge of reforming vasculature as well as throughout the more posterior inner retina where 5′N activity also was elevated. AI was also in intravitreal neovascularization.

conclusions. Peak adenosine levels in the inner retina correlated temporally with active vasculogenesis. Adenosine and 5′N levels were reduced in hyperoxia and then returned to above normal levels during the vasoproliferative stage of OIR.

Retinopathy of prematurity (ROP) is a potentially blinding disease in which degree of immaturity and, consequently, the susceptibility of developing retinal vessels to oxygen-induced vaso-obliteration are considered to be significant factors. ¹ Previous studies have demonstrated that the neonatal dog is a faithful model of human ROP and is the only model to have exhibited a cicatricial form of the disease. ^{2 3} The retina of the newborn beagle is only 60% vascularized at full-term birth and is developmentally similar to that of the human fetus at approximately 28 weeks gestation. ⁴ In the normal dog, formation of the primary or inner vascular network proceeds radially by a process involving in situ differentiation of angioblasts that subsequently coalesce to form primordial vessels (vasculogenesis). ^{4 5} During development, retinal Muller cells appear to play an integral morphologic role in primary vascular formation. Muller cell inner processes furnish angioblasts with an extensive network of glycosaminoglycan-rich extracellular spaces and provide a scaffold for pseudopodial attachment during angioblast migration and organization. ⁵  

Enzyme and immunohistochemical studies have demonstrated that the glycoprotein 5′nucleotidase (5′N) is localized in certain domains of Muller cells in several adult mammalian species ⁶ and during development in murine retina. ⁷ 5′N (EC 3.1.3.5) is an ecto-enzyme, which catalyzes the hydrolysis of purine nucleotide monophosphates, not pyrimidines, to their corresponding nucleosides. Although 5′N can metabolize all purine monophosphates, in ischemic rat hearts the major product is adenosine. ⁸ Braun et al. ⁹ recently demonstrated that 5′N expression is elevated during cerebral ischemia. In heart, 5′N is upregulated during hypoxia, and subsequently adenosine levels increase 50-fold. ¹⁰  

Adenosine, the major product of 5′N, has been proposed as an intercellular communication molecule, a modulator of synaptic transmission in brain ^{11 12} and in retina, ¹³ and a local regulator of blood flow in several organs. ^{14 15 16} In retina, adenosine modulates blood flow in adult and neonates ^{17 18 19} and is released in response to ischemia. ^{20 21 22} Dusseau and Hutchins ²³ demonstrated that hypoxia-induced angiogenesis on the chorioallantoic membrane (CAM) was due to adenosine production and uptake. In vitro, adenosine is chemotactic and mitogenic for endothelial cells from large blood vessels. ^{24 25} We have determined that adenosine does not stimulate proliferation of dog retinal microvascular endothelial cells but does stimulate endothelial cell migration and tube formation, two events that are critical in the development of the primary retinal vasculature in dog. ²⁶ Histochemical studies have demonstrated adenosine immunoreactivity (AI) in adult retinal neurons of several species of mammals. ^{27 28} Little is known, however, regarding either 5′N or adenosine distribution during the developmental period nor during oxygen-induced retinopathy. Considering the integral morphologic role of Muller cells during primary vasculogenesis and their capacity for producing vasogenic adenosine via 5′N, we examined 5′N and adenosine distribution during normal development of the retinal vasculature and during oxygen-induced retinopathy in the dog. 

Methods

Triplicate pure bred beagles were killed at 1, 5, 8, 15, and 22 days of age by an overdose of intraperitoneal sodium pentabarbitol. To produce oxygen-induced retinopathy (OIR), 1-day-old animals were placed in 95% to 100% oxygen for 4 days and then returned to room air as previously described. ^{29 30} Triplicate oxygen-treated animals were killed in 100% oxygen at 5 days of age to examine the vaso-obliterative phase, and then at 3, 10, and 17 days after return to room air to examine the vasoproliferative stage of the disease. Animals were handled in accordance with the tenets of the ARVO Statement for Use of Animals in Ophthalmic and Vision Research. One eye from each animal was slit at the limbus and frozen in OCT compound (Miles Scientific, Elkhart, IN) suspended in isopentane chilled with dry ice. Frozen blocks were stored at −70°C before sectioning. Twelve-micrometer cryosections were cut at −20°C, placed on glass slides, and air dried. From the fellow eye of each animal, the retina was incubated for ADPase and flat-embedded; results from these fellow eyes have been previously reported by McLeod et al. ^{3 29 30} Which eye was snap frozen for histochemistry was randomized and should not be of consequence since severity of retinopathy in the dog is remarkably bilateral. ³  

Enzyme Histochemistry

5′N activity was demonstrated using the technique of Wachstein and Meisel. ³¹ Sections were fixed for 5 minutes in 10% neutral buffered formalin at 4°C and then rinsed in three changes of distilled H₂O. Incubation was carried out for 30 minutes at 37°C in a medium containing 1.4 mM adenosine 5′-monophosphate (5′-AMP), 3.0 mM lead nitrate, 7.0 mM manganese chloride, and 0.2 M Tris maleate buffer at pH 7.2. After incubation, sections were rinsed in several changes of distilled H₂O and developed in 0.1% ammonium sulfide. Control incubations were performed in the absence of 5′-AMP. The specificity of the 5′N activity was demonstrated by including 1 to 100μ M α,β-methylene adenosine 5′-diphosphate (AMPCP), a well-characterized and specific inhibitor of 5′N, ^{32 33} in the incubation solution. It is important to note that this is not the enzyme activity that Irons ³⁴ described in photoreceptors, which uses pyrimidine not purine monophosphates as substrates at acid pH and was recently referred to as 5′ nucleotidase in a study concerning retinal detachment. ³⁵  

Menadione-dependent α-glycerophosphate dehydrogenase activity (M-α-GPDH) was localized as previously reported. ³⁶ In that study, we demonstrated that angioblasts and immature endothelial cells have the greatest levels of this enzyme, which declines with maturation. ³⁶ The form of α-GPDH associated with angioblasts was dependent on menadione, a derivative of vitamin K, and therefore represented the mitochondrial half of the glycerophosphate shuttle for bringing reducing equivalents into mitochondria. 

Immunohistochemistry

Vascular endothelial cells were labeled with an antibody against von Willebrand factor (vWf, diluted 1:20,000; Accurate Chemical Corp., Westbury, NY) using the immunohistochemical technique we have previously reported. ³⁷ After overnight incubation at 4°C with primary antibody, the sections were washed in phosphate-buffered saline (PBS) and then biotinylated goat anti-rabbit IgG (1:500; Kirkegaard and Perry, Gaithersburg, MD) was applied for 30 minutes. The sections were incubated with streptavidin labeled with peroxidase (1:500; Kirkegaard and Perry) for 45 minutes and were developed using 3-amino-9-ethylcarbazole as chromagen. 

Adenosine was detected immunohistochemically using anti-adenosine (anti-ADO) antiserum graciously provided by Andrew Newby, Bristol Heart Institute, University of Bristol, Bristol Royal Infirmary, UK. This rabbit antiserum was raised against adenosine conjugated to laevulinic acid (O2′,3′-adenosine-acetal). ³⁸ This antiserum has been used to detect adenosine in adult rat, guinea pig, monkey, and human retinas by Braas et al. ^{27 39} and mouse, rabbit, and ground squirrel by Blazynski et al. ²⁸ Binding in those studies was blocked by preincubation of antibody with 10 μM adenosine before use. The antibody or control rabbit serum was used at 1:10,000 with the following modification of the Braas’ technique. Sections were fixed in 2% glutaraldehyde, washed in PBS with 0.2% Triton X-100, blocked with 1:200 goat serum, and incubated at 4°C for 48 hours with primary antiserum or nonimmune antiserum. Sections were then washed with 0.05% Triton X-100 in PBS, blocked with 1:200 goat serum, and then incubated with goat anti-rabbit biotin for 90 minutes. The rest of the technique is the same as given above for vWf. All reagents were purchased from Sigma (St. Louis, MO) unless stated otherwise. 

Microdensitometric Analysis

Serial cryosections were processed in the following series, and the sequence was repeated three times for each eye: anti-vWf as a marker for differentiated endothelial cells; histochemical localization of M-α-GPDH activity for labeling angioblasts and immature endothelial cells ³⁶ ; 5′N enzyme histochemistry; and finally anti-ADO immunohistochemistry. ²⁷ Microdensitometry was performed on triplicate noncounterstained sections from each animal, from ora serrata to 7 mm posterior. Three separate measurements were made every millimeter in the inner retina (a total of 216 measurements in triplicate animals). Therefore, three measurements were made every 1 mm from ora serrata in triplicate slides from each animal, and there were three animals in each group. The SD from the triplicate readings in each area on the three slides from each animal are included in the resultant graphs to show reproducibility of the technique and geographic variation in an animal, since there was 100 μm distance in the tissue between each adenosine or 5′N triplicate slide. Direct microdensitometric comparisons were made on all sections from an age-matched control and oxygen-treated animal that were incubated at the same time in the same reagents so that immunohistochemical conditions were identical. It was not possible to make comparisons between all animals in the groups because of the variability in location of the edge of the vasculature in each animal, especially oxygen-treated animals, ³ even though the superior lobe was always used for analysis. 

The reaction products from 5′N enzyme histochemistry and anti-ADO immunohistochemistry were quantified using digitized images collected from a photomicroscope (Photomic II; Carl Zeiss, Thornwood, NY) equipped with a charge-coupled device (CCD) camera (Hamamatsu City, Japan) and a Macintosh IIci computer (Cupertino, CA) with a Data Translation frame grabber board and NIH Image software (version 1.47). The Zeiss Photomic microscope was alligned for Kohler illumination before performing the measurements. The sample with the most reaction product (22-day-old OIR animal for 5′N and anti-ADO) was used to set the gain and offset on the video system. The background was set for zero on the grayscale (central vitreous cavity for anti-ADO and 5′N). The darkest structure in the nerve fiber layer (Muller cell processes for 5′N and inner retina for anti-ADO) was used to set the values nearest 255 (upper limit of grayscale). This assured us that all density measurements were in the range of 0 to 255 arbitrary units (histogram optimization). Once the illumination, gain, and offset were set for each type of specimen, all images were captured under identical conditions. Density plot profiles were generated using rectangular field selections through the inner retina: 75 × 150 μm for adenosine; 20 × 150 μm for 5′N. The background density of the vitreous was subtracted from the peak density of each plot, which coincided with inner Muller cell processes for 5′N. The mean density and SD were calculated for each region or structure from a minimum of three density plot profiles, and statistical analysis of the data was performed using the two-tailed Student’s t-test with equal variance. 

Results

Vasculogenesis at 1 Day of Age

A representative group of serial sections from a 1-day-old animal is shown in Figure 1 . VWf immunoreactivity was present only in formed vessels (Fig. 1A) . M-α-GPDH–positive cells, on the other hand, were present in advance of formed vessels in the nerve fiber layer (Fig. 1B) . 5′N activity was greatest in the inner radial processes of the Muller cells, which provide the milieu in which angioblasts differentiate and organize. ⁵ Inclusion of 100 μM AMPCP, a well-characterized inhibitor of 5′N, in the incubation medium or exclusion of the substrate (AMP) completely eliminated the enzyme histochemical reaction product. There appeared to be a radial gradient of 5′N reaction product localized within the inner Muller cell processes of the nerve fiber layer; reaction product diminished in the Muller cell processes anterior to the leading edge of formed primordial vessels and was absent in the far periphery. Microdensitometry (Fig. 1E) confirmed that 5′N in inner retina was highest near and posterior to the edge of the forming vasculature and that it diminished peripherally. 5′N also was present at lower levels diffusely throughout the neuroblastic layer (Fig. 1C) . 

AI was most prominent in the nerve fiber layer and, to a lesser extent, diffusely throughout the neuroblastic layer (Fig. 1D) . Microdensitometry demonstrated that the peak of AI coincided with the edge of the formed vasculature (Fig. 1F) , as determined from serial sections incubated with anti-vWf. Microdensitometry also showed the dose dependence and reproducibility of the adenosine antibody binding in the 1-day-old dog (results not shown). The triplicate readings in each area yielded highly reproducible results, as indicated by the SDs, for relative reaction product density and for the degree reaction product increased as antibody titer became more concentrated (results not shown). 

Vaso-Obliteration during OIR

Analysis identical with that just described in the 1-day-old animal was done on animals exposed to 4 days of 100% oxygen starting at 1 day of age (Fig. 2) . The vasculature in the oxygen-treated animal was highly constricted and very few vWf-positive blood vessels remained (Fig. 2B) . ²⁹ The 5′N enzyme histochemical reaction product was greatly reduced throughout retina in the 5-day-old animals killed in oxygen compared to the room air controls (Figs. 2C 2D) . Microdensitometry demonstrated that the reduction of 5′N reaction product in inner retina was significant in all areas analyzed in this representative animal (Fig. 3A ). Adenosine reaction product also was greatly reduced in the inner retina of a 5-day-old animal killed in oxygen compared with its room air control littermate, presumably due to the reduction in 5′N activity after exposure to hyperoxia (Figs. 2E 2F 3B) . Therefore, substantial reduction in the vasodilator adenosine was accompanied by severe vasoconstriction and vaso-obliteration. Only a representative pair of animals is shown in Figures 3A and 3B and other graphs of densitometric values when room air control and oxygen-treated animals are compared because of the differences in distance between the ora serrata and the edge of the vasculature, especially those exposed to oxygen. However, the shapes of the curves and trends in reaction product densities were similar for the triplicate animals in each group. 

Vasoproliferative Stage in OIR

At 8 days of age, 5′N activity was still greatest in inner retina (Fig. 4C ) in room air control animals, although it was difficult to say at this age or older that it was confined exclusively to Muller cells. The greatest activity was near the edge of the developing vasculature, and activity diminished toward ora serrata in the representative animal shown in Figure 5A . In the oxygen-exposed 8-day-old animal, 5′N activity was greatly elevated throughout retina compared to control animals (Fig. 4D) , and the activity was uniform throughout the peripheral 7 mm of inner retina (Fig. 5A) . AI was present throughout the inner half of the retina in room air control animals (Fig. 4E) , but the peak of immunoreactivity in inner retina still coincided with the edge of the forming vasculature (Fig. 5B) . In oxygen-treated animals, AI was still greatest in the nerve fiber layer and present only in much lower levels in the remainder of the retina (Fig. 4F) . Microdensitometry demonstrated that the AI was greatest at and in advance of the edge of the vasculature in the oxygen-treated animal, and levels were greater than in the room air control animal except in far peripheral retina where vasculature was developing in the control animal (Fig. 5B) . 

At 15 days of age, the primary retinal vasculature had almost reached the ora serrata in the room air controls. Intravitreal neovascularization was prominent in the oxygen-treated animals, and most formations appeared immature morphologically as we have reported previously (Fig. 6B ). ^{3 30} 5′N activity was greatly elevated throughout retina in the oxygen-treated animals compared to room air controls (Figs. 6C 6D 7A ), but activity was not present in the intravitreal neovascularization. AI was significantly elevated in inner retina of the oxygen-treated animals compared to the room air controls, but the retinal vasculature had not progressed significantly toward the periphery (Fig. 7B) . Furthermore, high levels of immunoreactive adenosine were present in intravitreal neovascularization and in vitreous (Fig. 6F) . 

In the 22-day-old room air control animal, the primary vasculature had advanced to the ora serrata, and a secondary capillary network was forming in the inner nuclear layer (Fig. 8A ). Intravitreal neovascularization was prominent in oxygen-treated animals and most formations appeared mature morphologically (Fig. 8B) . ³ 5′N activity in room air control animals was no longer most prominent in the nerve fiber layer; the greatest activity was present in the inner plexiform layer (Fig. 8C) . 5′N activity was still greatly elevated in the oxygen-treated animals compared to controls, and it was present in glial processes at the base of intravitreal neovascularization (Fig. 8D) . Microdensitometric analysis demonstrated how substantial the difference was in 5′N activity in inner retina between the two groups (Fig. 9A ). Adenosine reaction product was greatly reduced in the nerve fiber layer in control animals (Fig. 8E) , appearing almost comparable in localization to that observed by Blazynski ²⁸ in adult retinas of several species. In oxygen-treated animals, adenosine reaction product was still greatly elevated in the entire inner retina (except at ora serrata; Fig. 9B ), where multiple layers of vessels formed, especially near the border of vascularized retina (results not shown). Adenosine reaction product was prominent in intravitreal neovascularization (Fig. 8F) . 

Summary of Results

The pattern of 5′N changed considerably as the retina developed (Fig. 10A ). At early ages, the greatest activity was in inner Muller cell processes (Figs. 1C 2C) . As development of the inner retinal vasculature reached completion at 22 days of age, activity in the inner retina decreased and activity in outer retina (both plexiform layers) increased (Fig. 8C) . At 1 and 5 days of age, the greatest AI was associated with inner retina (Figs. 1D 2E) , and the peak of immunoreactivity in inner retina was at the edge of formed vasculature (Fig. 10B) . The peak of adenosine reaction product shifted toward ora serrata as vascular development progressed radially and was coincident with the zone of active vasculogenesis. At 8 days of age, 5′N activity was still greatest in inner retina (Fig. 4C) in room air control animals, although it was difficult to say at this age and older in 12-μm cryosections that the activity was confined to inner Muller cell processes. By 15 days of age, the neuroblastic layer had differentiated into inner and outer nuclear layers, and the greatest activity was associated with the nerve fiber layer and the inner nuclear layer (Fig. 6) . By 22 days of age when radial progression of the inner retinal vasculature was complete, the pattern for immunoreactive adenosine was similar to that observed by Blazynski ²⁸ in adult mouse and rabbit. Adenosine reaction product and 5′N activity had decreased dramatically throughout the nerve fiber layer (Figs. 10A 10B) . Ganglion cells had the greatest adenosine reaction product, and inner nuclear layer and photoreceptor inner segments were also immunoreactive (Fig. 8E) . 

Hyperoxia greatly reduced both 5′N activity and AI in the 5-day-old retina. In general, there was a significantly elevated 5′N and adenosine reaction product in inner retina in the oxygen-treated animals when animals were returned to room air, i.e., the vasoproliferative stage of OIR. AI was also very high in neovascular formations that appeared in 15- and 22-day-old, oxygen-teated animals. 

Discussion

Elevated levels of adenosine at most ages were accompanied by elevated 5′N activity, and when 5′N activity shifted from the nerve fiber layer to the inner plexiform layer during development, AI also shifted. Similarly, levels and localization of 5′N and adenosine reaction product were comparable in OIR. Therefore, this study establishes an association between 5′N activity and adenosine levels and suggests that adenosine produced by Muller cell 5′N is involved in both retinal vasculogenesis and angiogenesis that occurs in the canine model of OIR. 

Normal Vasculogenesis

Although Kreutzberg and associates ⁶ previously demonstrated that 5′N was located in specific domains of the Muller cell in adult tissue, ours is the first report of 5′N being expressed in one domain of Muller cells (innermost processes) during development and then activity shifting to other domains in retina once vascular development is completed. Braun and associates ⁷ observed a shift in 5′N in developing mouse retina that they did not attribute to a change in Muller cell domain. The transient expression of 5′N activity within the inner Muller cell processes during the period of primary retinal vessel formation lends credence to the concept that Muller cells are intimately involved in vasculogenesis. Not only do these cells provide the glycosaminoglycan-rich extracellular milieu for angioblast differentiation, ⁵ but they also provide a stimulus for vasculogenesis, adenosine. 5′N catalyzes the final step in the conversion of intracellular 5′-AMP to membrane-permeable adenosine, which subsequently can be released into the extracellular space. Once released, adenosine may act on cells via surface receptors. There are two major classes of adenosine receptors in neuronal tissue, A₁ and A₂ receptors. In general, the A₁ receptors have been associated with neuronal elements and A₂ with vascular elements. ^{12 40} In Taomoto et al., ⁴¹ a companion article, we describe the localization of one subtype of adenosine receptor, A_2a, during retinal development and OIR. 

Teuscher et al. ²⁵ demonstrated in vitro that adenosine was chemotactic for porcine endothelial cells, i.e., stimulated migration in the direction of the highest concentration of adenosine. We have observed adenosine to stimulate chemokinesis, random movement of canine retinal microvascular endothelial cells, and also formation of cords and tubes in a collagen gel. ²⁶ Assuming that the radial gradient of immunoreactivity is indicative of a concentration gradient of adenosine (as demonstrated immunohistochemically herein), then Muller cells and the adenosine they produce may be paramount in normal retinal vasculogenesis and its progression toward the ora serrata. 

When vasculogenesis reaches completion and neuronal development progresses, as in the 22-day-old animals, 5′N activity and AI shift away from the nerve fiber layer. In older animals, 5′N was greatest in both plexiform layers, the synaptic region. Interestingly, the AI shifted to neuronal perikarya of ganglion cells and cells in the inner nuclear layer. The appearance of the adenosine reaction product in young versus old animals also was interesting. In the young animals, reaction product appeared diffuse in the inner retina, whereas in the 22-day-old, it was succinctly localized to ganglion cell perikaryon, as has been observed previously by Braas et al. ²⁷ and Blazynski ⁴⁰ in adult retinas. This suggests that the adenosine antibody may recognize both intra- and extracellular adenosine. 

At most ages in regions of retina analyzed, areas with high levels of 5′N activity had high AI. In some central areas of retina on days 1 and 5, however, there were high levels of 5′N, but relative adenosine reaction product densities were lower than at the edge of the vasculature. There are several possible explanations for this. A₁ receptor production by neuronal cells may increase as neurons differentiate and synaptic connections are established. If that were true, the demand for adenosine would be increased in that both neuronal and vascular elements would use adenosine. This would result in less available adenosine, even if adenosine production by 5′N did not change. Alternatively, if adenosine deaminase levels were increased, degradation of adenosine would occur even if 5′N activity remained the same. Adenosine deaminase has been demonstrated in retina, but changes in its level during retinal development have not been studied. ⁴² Adenosine also could be consumed through the salvage pathway by adenosine kinase. Gidday and Park ¹⁹ reported prominent vasodilation after intravitreal administration of an adenosine kinase inhibitor, strongly suggesting that adenosine kinase exists in neonatal retina. When adenosine is high and 5′N levels low (e.g., 22 days old), other adenosine-generating enzymes like S-adenosylhomocysteine hydrolase (SAH) may be producing adenosine. There have been no studies done to determine the levels of SAH in retina. 

Oxygen-Induced Retinopathy

During the vaso-obliteration stage of OIR in dog, there is a greater than fourfold decrease in the percentage of vascular area in retina. ²⁹ It is noteworthy that this is accompanied by a striking loss of 5′N activity and AI. The reduction in 5′N and adenosine may be due to oxygen radical damage to the enzyme during exposure to hyperoxia. This already has been demonstrated by Kitakaze and coworkers ^{43 44} in heart and in polymorphonuclear leukocytes. It is also possible that Muller cells in developing retina act as sensors of O₂ levels. During the initial hyperoxic environment in OIR, Muller cells may downregulate 5′N and, therefore, lower adenosine-favoring vasoconstriction, which occurs during the first 24 hours of exposure to hyperoxia. Patz, ⁴⁶ Ashton et al., ¹ and more recently, Chan–Ling et al., ⁴⁵ have suggested that the retina is in a state of physiological hypoxia after vaso-obliteration and during retinal vascular development. In retina, adenosine levels increase after induction of ischemia in other models, ²¹ suggesting that Muller cells could upregulate 5′N production in ischemic environs such as the retina during the vasoproliferative stage of OIR. In brain, hypoxia results in upregulation of glial 5′N. ⁹ In heart, 5′N is upregulated, and subsequently adenosine levels increase 50-fold during hypoxia. ¹⁰ It would therefore be logical that 5′N activity would be high as we have shown during vascular development and during the period of normoxia that follows a hyperoxic insult. Furthermore, the levels of immunoreactive adenosine were elevated at the same time as 5′N activities increased, as Kitakaze et al. ⁴⁴ have shown in heart. If Muller cells were capable of sensing oxygen levels, upregulation of 5′N in adult retina would serve to generate adenosine, which might function, as Braun et al. ⁹ have suggested, as a neuroprotectant in ischemic brain. Indeed, Roth et al. ²¹ have demonstrated that adenosine levels in retina increase significantly after experimental ischemic insult in adult retina. 

Concurrent with the elevation in adenosine after oxygen-treated animals are returned to room air is the angiogenesis that characterizes the vasoproliferative stage of OIR. Although we have not observed adenosine to stimulate proliferation of adult canine retinal microvascular endothelial cells, we have demonstrated that migration of these endothelial cells and organization into tubes is stimulated by adenosine ²⁶ and these represent two stages in angiogenesis. ⁴⁷  

There has been considerable attention paid to peptide growth factors, specifically vascular endothelial growth factor (VEGF), during development and angiogenesis in OIR in other species. ^{45 48 49 50} It has been suggested that hypoxia upregulates Muller cell expression of VEGF. ^{48 50} Adenosine, however, may control the production of VEGF in these events. It has recently been demonstrated that adenosine can stimulate production of VEGF. ⁵¹ Hypoxia-induced VEGF expression can be blocked by adenosine deaminase or adenosine receptor antagonists. ^{52 53 54} Furthermore, Muller cells produce both VEGF ⁴⁵ and adenosine. 

The concurrent shift of 5′N activity and AI suggests that 5′N may be the major source of adenosine in developing retina. The relationship of increased adenosine to retinal vasculogenesis and angiogenesis during OIR suggests that adenosine may stimulate these processes. This study suggests that production of adenosine and/or blockade of its receptors on vascular cells could be a therapeutic target for controlling angiogenesis during OIR. 

 Supported by National Institutes of Health Grants EY 01765 (Wilmer Institute) and EY09357 (GL) and the ROPARD Foundation. Gerard A. Lutty is an American Heart Association Established Investigator and the recipient of a Research to Prevent Blindness Lew Wasserman Merit Award.

 Submitted for publication November 24, 1998; revised June 28, 1999; accepted August 17, 1999.

 Commercial relationships policy: N.

 

Figure 1.

 

Relationship of developing retinal blood vessels to 5′N expression and AI in a normal 1-day-old dog. In serial sections shown in (A) through (D), the long thin arrows indicate the edge of the vasculature. In (A) and (B), the open arrows point toward ora serrata. The inner plexiform layer (ipl) is labeled in (A) for orientation. (A) Peripheral edge of canalized retinal vessels (arrow) as demonstrated by vWf immunoreactivity. (B) Same field in an adjacent section stained for M-α-GPDH demonstrates edge of formed vessels (arrow) and peripheral angioblasts (arrowheads). (C) 5′N activity was associated with inner Muller cell processes adjacent to developing blood vessels (arrow) and peripheral to the border of vascularized retina. (D) AI also was high in the inner retina at the edge of (arrow) and peripheral to the vascular border. Magnification, (A) through (D) ×50. (E) Microdensitometric analysis of the relative amount and distribution of 5′N activity in the inner retina in relation to the border of vascularized retina. The data represent the average for three animals, and the average edge of the vasculature in the 3 animals is indicated. (F) Microdensitometric analysis of the relative amount and distribution of AI shows the peak density near the vascular border. The data represent the average grayscale value for 3 animals, and the average edge of the vasculature in the three animals is indicated. (G) High-magnification micrograph of 5′N activity in inner Muller cell processes just anterior to the border of vascularized retina. Open rectangles indicate examples of profile plot selections used for microdensitometric analysis of Muller cell–associated 5′N reaction product. (H) Glycol methacrylate section (2.5 μm thick) from ADPase flat-embedded fellow retina to that shown in (G) demonstrates the structure of inner Muller cell processes at high resolution (paired arrows). These structures shown anterior to the border of vascularized retina are identical with those intensely labeled structures shown in (G). Magnification, (G, H) ×470.

Figure 2.

 

Blood vessels (vWf), 5′N and adenosine in a 5-day-old room air control (A, C, E) and in a 5-day-old animal killed after 4 days in oxygen (B, D, F). Fields shown in all plates are 4 to 5 mm from the ora serrata. The open arrows in the top panels point in the direction of ora serrata. The inner plexiform layer (ipl) is labeled only in the top panels for orientation. (A) VWf immunohistochemical labeling of blood vessels just posterior to the border of vascularized retina demonstrates the dilated developing primary vasculature in the inner retina of the room air control (solid arrow). (B) Four days of hyperoxia results in the obliteration of most blood vessels and extreme constriction of the few remaining viable vascular channels (solid arrow). (C) 5′N activity was greatest in the inner Muller cell processes, which surround the normal developing vessels (solid arrow). (D) There was a significant decrease in 5′N in the inner Muller cell processes in all regions analyzed of the oxygen-treated animals. (E) AI was highest around developing blood vessels in the inner retina of the normal 5-day-old room air control (solid arrow). (F) Like 5′N, adenosine was much less prominent in the inner retina of animals after prolonged oxygen breathing. Magnification, ×50.

Figure 3.

 

Plots showing the density and distribution of inner retinal 5′N activity (A) and AI (B) in a representative 5-day-old air-control animal (•) and a 5-day-old animal after 4 days of hyperoxia (□). Arrows indicate the edge of the vasculature in both animals. Analysis revealed a significant decrease in both 5′N activity and AI throughout the inner retina of oxygen-treated animals compared to the room air controls. Error bars represent the mean grayscale values for triplicate readings in triplicate slides at that site, and, therefore, simply demonstrate the reproducibility of the technique.

Figure 4.

 

Blood vessels, 5′N activity and AI in an 8-day-old room air control animal (A, C, E) and an 8-day-old, oxygen-treated animal three days after return to room air (B, D, F). The open arrows in the top panels point in the direction of ora serrata for all micrographs. (A) Area of retina just posterior to the edge of forming vasculature showing dilated capillaries immunoreactive for vWf (short bold arrow). (B) Area of posterior retina 5 to 6 mm from ora serrata showing the edge of reforming vasculature after hyperoxic insult (long thin arrow). (C) Same area shown in (A) in a serial section showing high 5′N activity in the inner retinal Muller cell processes adjacent to developing vessels of the room air control animal (short bold arrow). (D) In the oxygen-treated animal, 5′N activity was increased at the edge of reforming vasculature (long thin arrow) and throughout the inner retina. (E) AI was highest around the blood vessels in the inner retina of the room air control (bold solid arrow). (F) In the oxygen-treated animal, AI was increased at and in advance of the edge of reforming vasculature (long thin arrow). Magnification, ×80.

Figure 5.

 

Plots showing the density and distribution of inner retinal 5′N activity (A) and AI (B) in a representative 8-day-old air control animal (•) and an 8-day-old, oxygen-treated animal after 3 days return to room air (□). Arrows indicate the edge of the vasculature in all plots. Microdensitometric analysis revealed a significant increase in 5′N activity throughout the inner retina of oxygen-treated animals compared to the room air controls (A). AI was highest at and in advance of both the normal developing vasculature in the room air control and the reforming vasculature of the oxygen-treated animal (B). There was a significant increase in AI in the posterior retina of this oxygen-treated animal (4–7 mm from ora) compared to the room air control. Error bars represent the mean grayscale values for triplicate readings in triplicate slides at that site, and, therefore, simply demonstrate the reproducibility of the technique.

Figure 6.

 

Blood vessels, 5′N activity and AI in a 15-day-old air-control animal (A, C, E) and a 15-day-old, oxygen-treated animal 10 days after return to room air (B, D, F). The open arrows in the top panels point in the direction of ora serrata in all micrographs below them. The inner plexiform layer (ipl) is labeled in the top micrographs for orientation in the serial sections below. The bold solid arrows point to the same retinal vessel in each animal. (A) Immunolabeling of retinal blood vessels for vWF (short bold arrow) just posterior to the border of vascularized retina in the room air control. (B) Retinal vessels (short bold arrow) and intravitreal neovascularization (long paired arrows) just posterior to the border of vascularized retina in the oxygen-treated animal. The vitreoretinal interface is indicated (v). (C) In the room air control, 5′N activity was still associated with the inner retina adjacent to the retinal vessels (short bold arrow) in the peripheral retina. (D) Compared to the room air control, 5′N was elevated in the inner retina of the oxygen-treated animal adjacent to blood vessels (short bold arrow) and was present in the glial processes at the base of intravitreal neovascularization (arrowhead). (E) Like 5′N, AI was highest in the inner retina adjacent to forming blood vessels (short bold arrow) of the peripheral retina of the room air control. (F) Compared to the control, AI was elevated in the inner retina adjacent to blood vessels in the oxygen-treated animal (short bold arrow) and was high in the intravitreal neovascular formation (long paired arrows). Magnification, ×80.

Figure 7.

 

Plots showing the density and distribution of inner retinal 5′N activity (A) and AI (B) in a representative 15-day-old room air control animal (•) and a 15-day-old, oxygen-treated animal after 10 days return to room air (□). Arrows indicate the edge of the vasculature in all plots. Microdensitometric analysis revealed a significant increase in 5′N activity in most regions of inner retina of the oxygen-treated animal (2 and 4–7 mm) compared to the room air control (A). AI was significantly elevated throughout the inner retina of the oxygen-treated animal except at the ora (B). Error bars represent the mean grayscale values for triplicate readings in triplicate slides at that site, and, therefore, simply demonstrate the reproducibility of the technique.

Figure 8.

 

Blood vessels, 5′N and adenosine in a 22-day-old room air control animal (A, C, E) and a 22-day-old, oxygen-treated animal 17 days after return to room air (B, D, F). In the top panels, the open arrows point in the direction of ora serrata. The vitreoretinal interface (v) and the inner plexiform layer are (ipl) are labeled for orientation. (A) In the control, the primary vascular network had reached the far periphery and the primary (short bold arrow) and secondary capillary network (below the ipl) was established posteriorly as shown in this vWf-immunolabeled section. (B) In the oxygen-treated animal, blood vessel growth toward the periphery was retarded, and intravitreal neovascularization was present posterior to the border of vascularized retina (long paired arrows). (C) With primary vasculogenesis complete and secondary capillaries formed, 5′N activity became less associated with inner Muller cell processes and most prominent in the inner plexiform layer (ipl). (D) In contrast, oxygen-treated 22-day-old animals exhibited high 5′N in all retinal layers with the most activity being localized to the inner retina. 5′N activity was not associated with intravitreal vessels but was localized to glial processes at the base of feeder vessels (arrowheads). (E) In the control, AI was localized to ganglion cell bodies in the inner retina (curved arrows) and neurons of the inner nuclear layer below the ipl. Photoreceptor inner segments (is) also were immunoreactive. (F) Like 5′N, adenosine was greatly elevated in the inner and outer retinal layers of the oxygen-treated animal. Intravitreal vessels were also immunoreactive (long paired arrows). Magnification, ×80.

Figure 9.

 

Plots showing the density and distribution of inner retinal 5′N activity (A) and AI (B) in a representative 22-day-old room air control animal (•) and a 22-day-old, oxygen-treated animal after 17 days return to room air (□). Arrows indicate the edge of the vasculature in all plots. Microdensitometric analysis revealed a significant increase in 5′N activity in the inner retina of the oxygen-treated animal compared to the room air control in all regions except the ora (A). AI was significantly elevated in most regions of inner retina (2–7 mm from ora) of the oxygen-treated animal (B). Error bars represent the mean grayscale values for triplicate readings in triplicate slides at that site, and, therefore, simply demonstrate the reproducibility of the technique.

Figure 10.

 

Comparison of density and distribution of 5′N (A) and adenosine (B) as a function of age showing high levels of both 5′N and adenosine in normal inner retina during vasculogenesis and a decline when vascular development is nearly complete (22 days of age). Arrows indicate the average edge of vasculature at each age for 3 animals. Data represent the average grayscale values for 3 animals in each group.

The authors thank Andrew Newby for providing the anti-adenosine antibody, without which this study could not have been performed.