A Novel Model of Retinopathy of Prematurity Simulating Preterm Oxygen Variability in the Rat

Abstract

purpose. To examine changes in the retinal vasculature of rat pups after 14 days of minute-by-minute small variations in oxygen.

methods. Arterial oxygen data from a preterm infant who developed severe retinopathy of prematurity (ROP) was translated to equivalent values for the rat. Newborn rat pups were raised for 14 days in a cage in which a computer controlled the atmosphere to mimic the fluctuating oxygen profile (group V). Positive controls (P) of 12-hour cycles of 80% and 21% were run concurrently, as were room air controls (C). All were killed at day 14.

results. Groups V and P had significantly larger avascular retinal areas than C[ median, interquartile range (IQR): 1.7%, 0–7.9%; 10%, 8.1–13%; 0%, 0–0%, respectively; each group n = 30]. Group P had a higher capillary branch count than C (median, IQR: 310/mm²; 253–311 mm²; versus 277/mm², 272–364/mm², respectively), but this was not significant using a multilevel analysis. Group V had significantly reduced capillary counts compared with C (median, 261/mm²; IQR, 215–290/mm²; P < 0.05 multilevel analysis). No neovascularization was seen in any group, though abnormal terminal vessels were seen at the avascular/vascular retina interface in 73% of rats in group P and 21% of rats in group V. In situ hybridization on serial sections demonstrated VEGF in the inner nuclear layer of the retina in P and V, whereas C showed trace levels only.

conclusions. The vaso-obliterative stage of ROP can be induced in rats using clinically relevant oxygen levels.

Retinopathy of prematurity (ROP) is a potentially blinding condition of preterm infants first described in 1942 in infants who were 8 weeks preterm. ¹ Retinopathy was associated with the introduction of high concentrations (80%) of supplemental oxygen given to preterm infants without arterial monitoring, for up to 8 weeks. A reduction in inspired oxygen concentration in subsequent decades reduced the incidence of the disease, though death and handicap from hypoxia increased. ² In the 1970s and 1980s, the development of techniques to continuously monitor oxygen (both with intra-arterial probes and transcutaneous probes), together with improved ventilators with better control of the inspired oxygen content enabled the arterial oxygen to be maintained within “safe” limits. However, the increased number of extremely immature infants (up to 16 weeks preterm) who are able to survive because of these and other technological advances may be the reason that the incidence of ROP has once again begun to increase. ³  

Although ROP is now considered by clinicians to be multifactorial, animal work has continued to identify oxygen as a factor in development. Animal work on oxygen-induced retinopathy (OIR) that represents ROP has predominantly been based on the extreme oxygen injury model that was responsible for retinopathy in relatively mature preterm infants in the 1940s. ^{4 5} More recently, Penn ⁶ and others ⁷ have varied the inspired oxygen delivered to models over 6- to 48-hour periods in an active endeavor to more closely resemble the arterial oxygen of preterm infants, though they recognize that this only partially mimics what is experienced by an extremely preterm infant. Relative hypoxia appears to be an important contributor to retinopathy, ⁸ and our understanding of the control of vascular development has increased with the identification of vascular endothelial growth factor (VEGF) and its hypoxic induction via hypoxia inducible factors (HIFs). ⁹ Recent preliminary work by Madan et al. ¹⁰ has suggested that at least one of the HIF proteins may be involved in the development of ROP. 

Practically, preterm infants have significant swings in their arterial oxygen though they stay predominantly within clinically acceptable“ safe limits.” Our previous work with a group of preterm infants identified that severe ROP was associated with greater variability in transcutaneous (arterialized) oxygen in the first 2 weeks of life. ¹¹ We hypothesize that these frequent small changes in oxygen may cause retinopathy by a frequent interruption of the process of ordered retinal vascularization and stabilization, which is controlled by relative hypoxia and hyperoxia in response to blood supply and metabolic demands. We have, therefore, developed an oxygen delivery system that enables us to mimic the transcutaneous arterialized oxygen values recorded from an infant who developed severe ROP. ¹² This is a more representative model of newborn arterial oxygen than has previously been possible. 

Methods

This was a group controlled study with masking of samples before assessment. Approval for this study was given by the UK Home Office, and all animals were cared for in accordance with UK Home Office legislation and the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Oxygen Profile Development and Delivery

A computerized physiological monitoring system has gathered data from infants admitted to the Neonatal Intensive Care in Edinburgh since 1990. ¹³ Arterial oxygen is continuously monitored transcutaneously and a mean of data points is stored each minute. The computerized transcutaneous oxygen profile of one infant who developed severe (threshold) ROP was selected, and the first 14 days were used. A stream of transcutaneous oxygen values was obtained, representing partial pressure of arterial oxygen, one value per minute. Published data were used to translate arterial oxygen in preterm humans ¹⁴ to arterial oxygen in the rat. In brief, a preterm infant in our unit is maintained between 6 and 10 kPa (45–75 mm Hg), with a mean of 8 kPa (60 mm Hg). A newborn rat breathing 21% oxygen has an arterial value of 12.9 kPa (96.8 mm Hg). Therefore to each arterial value gained from the preterm infant we added 4.9 kPa (36.8 mm Hg) to give an equivalent value in the rat. From this set of values we derived the inspired oxygen in the rat that would produce the equivalent arterial oxygen, ¹⁵ one value per minute (Fig. 1) . 

A computer-controlled delivery system was devised that was capable of changing the oxygen concentration within a small animal isolator within 1 minute (Reming Bioinstruments Co., Redfield, NY). This system is described elsewhere. ¹² Briefly, the system delivers either oxygen or nitrogen to effect the required atmospheric change in oxygen. It is capable of producing up to a ±50% change in atmospheric oxygen within 1 minute and proved at testing to be sufficiently precise. The median difference between required and monitored value of oxygen was 0.3%, with an interquartile range (IQR) of 0.2–0.7% (n = 17,465). Eighty-five percent of all monitored readings were within 1% oxygen either side of the set-point; 95% were within ±2% oxygen. 

Animal Groups

Three groups of animals were studied, each for 14 days. A control group (C) consisted of animals raised in room air. A 12-hour-variable model (P) consisted of a group raised in oxygen fluctuating between 80% and 21% (room air) on 12-hour cycles. Our experimental model, the minute-variable group (V), were animals exposed to our oxygen profile, with changes in inspired oxygen concentration each minute. 

Pregnant animals were acclimatized to isolators at least 24 hours before delivery. A minimum of 12 pups was required per litter, which all rat mothers produced in these experiments. Experiments were begun immediately after the delivery of the final pup in each litter. Bedding was changed every 7 days, at which point the profile was paused and restarted a few minutes later. No other interruption to the profile was required. 

At 14 days the rat pups were weighed and anesthetized by intraperitoneal injection of ketamine (2.5 mg/kg) and xylazine (1 mg/kg). Paraformaldehyde (PFA) was then directly perfused (0.4 ml 0.5%) into the left ventricle, and then pups were euthanatized by intracardiac injection of pentobarbitone (80 mg/kg). Both eyes were enucleated. 

Preparation of Tissues

Retinal Wholemounts.

The retinas were dissected using a modification of the method of Chan-Ling. ¹⁶ Enucleated eyeballs were fixed whole in 2% PFA for 2 hours before being washed in 1 M phosphate-buffered saline (PBS), pH 7.4. Under a dissecting microscope an incision was made between the cornea and sclera. Scissors were then used to cut around the junction between the cornea and sclera until the cornea could be removed. The lens was gently removed, taking care not to remove the retina. The eyecup was transferred to 1 M PBS for further dissection. The retina was gently eased from the sclera using fine forceps, taking care to leave the ora serrata intact because it defines the edge of the retina. The retina was then placed onto a TESPA (3′-aminopropyltriethoxysilane)-coated slide and flattened by making four or five incisions perpendicular to its outer edge. At this stage as much vitreous as possible was removed using cellulose sponges and scissors. 

The flattened, wholemounted retinas were permeabilized in 70% vol/vol ethanol (kept at −20°C) for 20 minutes and then in 1 M PBS/1% Triton X-100 for 30 minutes. The retinas were then incubated with biotinylated Griffonia simplicifolia (Bandeiraea) isolectin B4 (ICN, Hampshire, UK) at 5 μg/ml in 1 M PBS overnight at 4°C. They were rinsed in 1 M PBS/1% Triton X-100 for 10 minutes and then twice in 1 M PBS for 10 minutes. Streptavidin-conjugated fluorescein isothiocyanate (FITC; Sigma, St. Louis, MO) at 25 μg/ml in 1 M PBS was added for 4 hours at room temperature, and the slides were rinsed twice in 1 M PBS for 10 minutes each. The retinas were mounted in PBS:glycerol (2:1), and the coverslip was sealed with nail varnish. 

The stained, flatmounted retinas were viewed using an argon krypton laser confocal microscope (Leica Microsystems GmBH, Heidelberg, Germany), which allowed low- and high-powered images to be taken and digitally stored for later analysis. 

Capillary Density.

Capillary bed sample areas were chosen in the central retina with no major vessels present in the fields analyzed. Five areas of capillary vasculature in each retina were imaged at ×100 magnification and stored for later analysis. All stored files were assigned a random number to mask the observers, and counts of the number of branches were made. One observer counted all files, and a second observer counted a subset to compare results. There was a statistically significant correlation (P < 0.05) between the two observers using a Pearson correlation and no evidence of bias assessed using a Bland-Altmann plot. 

Avascular Areas

Digitized images of the total retinal area and peripheral avascular area were measured using Scion Image Software (Scion Corporation, Frederick, MD), and the avascular area was expressed as a percentage of the total retinal area. 

Immunohistochemistry

The whole enucleated eyeball was pierced by a needle and fixed in 4% PFA for 2 hours. It was washed twice in 1 M PBS (pH 7.4) before being embedded in agarose (1.5% in 1 M PBS, pH 7.4, supplemented with 5% sucrose). The solid agarose blocks were trimmed and left in 30% sucrose overnight at 4°C or until the agarose block sank. The blocks were then frozen slowly and stored at −70°C until sectioning. Serial 10-μm-thick cryosections were made from whole frozen eyes and incubated for 1 hour using biotinylated G. simplicifolia (Bandeiraea) isolectin B4 (ICN) at 12.5 μg/ml in Tris-buffered saline (TBS). After successive washes in TBS, the slides were then incubated with peroxidase-labeled streptavidin (Dako, High Wycombe, UK) at 8.75μ g/ml for 1 hour. After washing, DAB was applied for 5 minutes, and the slides were rinsed in tap water and counterstained in hematoxylin before being mounted in Depex (BDH Chemicals, Poole, UK) and viewed under a light microscope. Slides were analyzed for preretinal vessels that grow out from the surface of the retina. 

Cryosections

Sections of 10 μm were cut from frozen eyes (as above) using a Leica CM 1300 cryostat and transferred to TESPA-coated slides. Slides were air-dried for at least 30 minutes before storage at −20°C until required. 

VEGF Probes.

DNA from exons 1 to 4 (392 bp) were cloned into a pBluescript II KS (Stratagene, La Jolla, CA) at the SacII site (gift from Steve Charnock-Jones, University of Cambridge, UK). The plasmid was digested with SacI restriction endonuclease and end-filled using T4 DNA polymerase. A DIG-labeled sense probe was generated using T3 RNA polymerase (Roche Molecular Biochemicals, East Sussex, UK). The antisense probe was generated by digesting the plasmid with BamHI restriction endonuclease. After purification using Elutip columns (Schleicher & Schuell, Keene, NH), the DNA was transcribed in vitro with T7 polymerase in the presence of DIG-UTP (Roche Molecular Biochemicals). The size of both probes was confirmed using agarose gel electrophoresis. 

Methods.

Frozen sections were allowed to defrost at room temperature for at least 1 hour and hybridized with probe overnight at 65°C. Slides were washed at 65°C in SSC buffer (3 M sodium chloride, 0.3 M sodium citrate, pH 7) with 50% formamide and 0.1% Triton X-100 and then at room temperature in TBST (0.14 M NaCl, 2.7 mM KCl, 0.025 M Tris-HCl, pH 7.5, 1% Triton X-100). Slides were then blocked in 10% heat-inactivated sheep serum (in TBST) for >l hour at room temperature. Anti-DIG AP-Fab fragment (in 10% heat-inactivated sheep serum in TBST) was then added to each section and incubated overnight in a humidified chamber at 4°C. Slides were washed in TBST at room temperature and then in NTMT (100 mM NaCl, 100 mM Tris-HCl, pH 9.5, 50 mM MgCl₂, 0.1% Triton X-100). Staining was performed in the dark with nitroblue tetrazolium (NBT + 3.5% 5-bromo-4-chloro-3-indolyl phosphate [BCIP] in NTMT. After a few hours, staining reaction was checked, though color development could take up to 24 hours at 4°C. The reaction was stopped using distilled water, and the sections fixed in 4% PFA/0.1% glutaraldehyde for 20 minutes. The sections were dehydrated through a series of alcohols and then counterstained with filtered 0.1% eosin in 95% ethanol for 20 to 30 seconds. Rinses were made in 95% and then in 100% ethanol before transferral to histoclear and mounting in Vecta mount (Vector Laboratories, Peterborough, UK). 

VEGF Concentration in Retinal Sections.

The presence or absence of VEGF mRNA assessed by in situ hybridization in retinal sections was qualitatively scored by two masked, independent observers. 

Statistics

Summary statistics are presented as means ± SD or median and interquartile range. Correlation between groups was by Pearson correlation and correlation of between-group differences was made by Mann–Whitney U. The capillary branching was compared between groups by multilevel analysis using the software package MLWin (Institute of Education, University of London). 

Results

Assessment of Vascular Injury

Three methods of assessing changes in the vasculature are described: two quantitative measures of the vasculature from an assessment of lectin-stained retinal flatmounts (radial extent of retinal vasculature and capillary concentration) and a semiquantitative assessment of retinal sections by in situ hybridization for VEGF. All samples were randomized once processed for analysis and counts. 

Retinal Wholemounts

Peripheral Avascularity.

Thirty animals were assessed in each of the three groups at 14 days postnatal age. Control animals had fully vascularized retinas; those exposed to 12-hour-variable oxygen injury had a median 10% peripheral avascular area, and the minute-variable oxygen had median 1.7% peripheral avascularity (Table 1) . The differences between groups were statistically significant in all cases (P < 0.001; Mann–Whitney U). The degree of peripheral avascularity in group P is consistent with that seen with other extreme oxygen injury models in the rat. Penn ¹⁷ produced 8% peripheral avascularity at 14 days after a 24-hour-variable 80/40% oxygen regimen but without neovascularization. ⁶  

Figures 2A 2B 2C shows lectin-stained retinal flatmounts demonstrating the typical pictures of the relative degree of avascularity in each of the three groups. 

Capillary Density.

Figures 2D 2E 2F shows images captured from one typical retinal capillary bed from each experimental group. Capillary density was found to be higher in group P, though the difference was not statistically significant when using a multilevel analysis. The vasculature in group P was qualitatively different from that of the controls or minute variables, the appearance of which suggested an immature capillary bed. Capillary-free zones are began to appear, suggesting that remodeling was starting to happen. There was significantly lower capillary density in the minute-variable group V compare with control (P < 0.05) using multilevel analysis. 

Abnormal Vessels

Observers noted no extraretinal neovascularization on the flatmounts, though these are often difficult to distinguish in these preparations. However, two masked observers noted abnormal terminal dilatations present at the vascular/avascular interface of 73% of retinas from group P and in 21% of retinas from group V. Figure 2J shows a typical example of one of these terminal dilatations from the group P. They were stained with endothelial cell–specific lectin and may represent endothelial cell proliferations. 

Assessment of VEGF by In Situ Hybridization

VEGF mRNA was demonstrated in retinas using in situ hybridization and scored semiquantitatively. Sense controls were included and showed no staining (results not shown). VEGF was found in all three groups in both the anterior retina and the retina as a whole, though staining was more intense in the anterior retina in most specimens. There was an increasing strength of staining present in group V compared with group C, and in group P when compared with group V. Figures 2G 2H 2I demonstrate a typical cryosection with VEGF mRNA stained in black from a retina in each of the three groups. VEGF is evident in the inner nuclear layer as previously reported. ¹⁸  

Immunohistochemistry

Immunohistochemistry was performed on serial cryosections from each group, but no evidence of extraretinal neovascularization was seen. 

Postexperimental Weights

Mean body weights at 14 days were as follows: group C, 29.2 g (95% confidence interval [CI], 28.3–30.2 g); group P, 28.6 g (CI, 27.8–29.5 g); and group V, 23.7 g (CI, 23.0–24.4 g). The difference between the variable group and other two was significant (P < 0.001; t-test), but there was no difference between group P and group C. 

Discussion

Our model has demonstrated that small frequent changes in inspired oxygen are able to induce retinopathy in a neonatal rat. Although previous groups have induced retinopathy using fluctuations of 40% to 70% oxygen ^{6 7 17} over periods of 6 to 48 hours, our model produced a change in oxygen each minute (median change, 0.8% oxygen; range, 0–21.8%). Our small but frequent changes in oxygen, mimicking the changes in arterial oxygen seen in a preterm infant, were able to significantly reduce both peripheral vascularity and the concentration of capillaries when compared with room air controls. We did not however, induce neovascularization but noted abnormal dilated terminal vessels at the vascular/avascular interface. In addition, VEGF mRNA was present in the inner nuclear layer of the retinas in our model, whereas room air controls had little if any VEGF present. 

The development of models for identifying the pathogenesis of ROP has predominantly concentrated on extreme oxygen injury and are based on early and pivotal work by Ashton ¹⁹ that first established that oxygen was important in disrupting retinal blood vessel development. Although these models are able to induce retinopathy with neovascularization, ^{4 5} they have not closely represented the arterial oxygen of the preterm infant who might in current times develop the disease, and this has been considered a potential weakness in understanding the precise pathogenesis of the disease in extremely preterm infants. Work by Penn ^{6 15} clearly delineated that the relative degree of hyperoxia and hypoxia were important in inducing OIR, ¹⁷ and his most successful model, which produced severe OIR with neovascularization in all retinas, used fluctuations between 10% and 50% inspired oxygen. Penn’s levels of oxygen mimic the extremes that a preterm neonate could experience and support the concept that our understanding of retinopathy may be enhanced by models more closely representing preterm arterial oxygen levels. We have extended that concept and modeled our fluctuations directly on those that were experienced by a preterm infant that developed severe ROP. The extremes of our oxygen fluctuations were similar to Penn’s (9.2–41.5%); however, each step change was relatively small but frequent. Although both our model and the extreme injury models are able to induce retinopathy, only the extreme injury models are successful in inducing the later, proliferative stages of the disease. 

The ability of small frequent fluctuations in oxygen to create retinopathy may be linked to the intrinsic nature of VEGF induction and suppression. VEGF has a relatively short half-life in normoxia (40 minutes) when compared with other mammalian mRNA but is stabilized during hypoxia. ²⁰ Although this enables a rapid response rate to stimulus, in the preterm infant, persistently fluctuating oxygen may make this system too responsive and inefficient in producing stable new blood vessels. In our model periods of relative hyperoxia alternate with periods of relative hypoxia (Fig. 1) , and presumably, stimulation would alternate rapidly with suppression. Pierce et al. ²¹ have demonstrated that hypoxic stimulation of VEGF can be reversed by a 24-hour exposure to hyperoxia, which reduced VEGF mRNA to undetectable levels and reduced VEGF protein to 70% of its hypoxic level. Their study did not assess the effects of shorter periods of hypoxia/hyperoxic stimulation and suppression of VEGF. We assessed VEGF mRNA in our model using in situ hybridization, which does not quantify the amount of VEGF mRNA present. However, in group V, VEGF mRNA was clearly present centrally within the inner nuclear layer and both peripherally and centrally throughout the retina. In contrast, room air controls had little or no VEGF mRNA present. 

Neovascularization was not produced in our model. However, in both our minute-variable and 12-hour-variable pups we did see abnormal terminal dilatations present at the vascular/avascular interface of the retina. These could either represent precursors of pathologic neovascularization or simply endothelial proliferation at the leading edge of retinal vasculogenesis. A period in room air has induced neovascularization in the 12-hour-variable model ⁶ and might have done so in our model. However, we chose not to have a room air exposure period at the end of the experiment in any of our experimental groups because preterm infants do not experience an acute switch from high to relatively low oxygen during which they develop neovascularization. They progress to threshold ROP while still experiencing variable arterial oxygen, probably as a result of more chronic lung disease. 

Although the lack of neovascularization in our minute-variable model may be seen as indicating a poor model of OIR, we believe that the pathophysiological process induced by our model may be more representative of the earlier stages of ROP than the extreme oxygen injury models. We ran an extreme oxygen injury model of 12-hour variations between 80% and 21% for 14 days as a comparison for our minute-variable model. In the 12-hour-variable group we demonstrated a large avascular peripheral retina and VEGF staining in the inner nuclear layer but no central vaso-obliteration. The larger vessels had capillary free zones (Fig. 2E) , but the capillaries between these areas were dilated compared with controls, and there were more capillary branches, which suggested that the vasculature may be immature. Other investigators have demonstrated occlusion and obliteration of capillaries both centrally and peripherally ^{15 17 22 23} using an extreme oxygen injury models in rats, and our inability to produce central vaso-obliteration is difficult to interpret. The experimental protocol we used was not identical with others, in that we used a combination of Reynauds ²³ 60%Δ Fio₂ and Penn’s ¹⁷ 12-hour cycles. 

Further work is needed with our model to assess the precise pathophysiology induced by frequent small fluctuations and to assess the degree of capillary obliteration and astrocyte injury. In addition, we need to further assess the influence of frequent oxygen fluctuation on vascular stabilization in our minute-variable group. 

Our novel animal model, based on the arterial oxygen data derived from an infant with ROP, is able to more closely represent the retinal oxygenation of a preterm infant developing ROP than has previously been possible. Though the retinopathy produced was not as severe as with other extreme injury models, perhaps this will enable us in the future to delineate a pathophysiological process in the disease that has not been represented by previous animal models. 

²  These authors should be considered as having contributed equally to this article.

 Supported by Royal Blind School, Edinburgh, United Kingdom; Mason Medical Foundation; Ross Foundation for the Prevention of Blindness; Research into Eye Disease Trust; and the Royal Society.

 Submitted for publication June 6, 2000; revised August 3, 2000; accepted August 17, 2000.

 Commercial relationships policy: N.

 

Figure 1.

 

Translated inspired oxygen values for the rat. The main graph is all of the 21,328 data points (14 days of minute-by-minute data). Insets: 6 hours of the whole profile from day 1 (a) and day 12 (b).

Table 1.

 

Analysis of Lectin-Stained Whole Flatmounted Retinas for Each Experimental Group

	No. of Branches/mm2	Peripheral Avascularity (% Total Retina Area)	Percentage of Retinas with Abnormal Terminal Vessels
Control (n = 30)	277	0	0
	(253–311)	(0–0)
Minute variable (n = 30)	261*	1.7**	73
	(215–290)	(0–7.9)
12-hour variable (n = 30)	310	10**	21
	(272–364)	(8.1–13)

 Values are median (interquartile range).

*  P < 0.05 multilevel analysis.

†  P < 0.001 Mann-Whitney U.

Figure 2.

 

(A through F, J) G. simplicifolia (Bandeiraea) isolectin B4–stained retinal wholemounts from 14-day-old rats. (G through I) cryosections through the center of the eye from 14-day-old rats, probed for VEGF mRNA and counterstained with eosin. (A, D, G) Room air controls (group C); (B, E, H, J) 12-hour-variable group (group P); (C, F, I) minute-variable goup (group V). (A through C) Edge of a retina taken from a representative image in each experimental group showing the degree of avascularity (magnification, ×10). (A) Room air controls are fully vascularized; (B) large avascular area; (C) avascular area smaller than that in (B) but still present. (D through F) An area of capillary bed from a retina in each experimental group (magnification, ×20). (D) Group C; (E) group P; white arrows, capillary-free areas, but the vasculature looks unremodeled; (F) group V; lower capillary density compared with (D) and (E); white arrows, isolated areas of lectin stain. (G through I) Cryosections through the center of the eye in one of each of the experimental groups (magnification, ×50); white arrows, retinal surface. (G) No staining; (H, I) strong staining throughout the inner nuclear layer. (J) Peripheral avascular area from one retina in group P, showing an abnormal terminal dilated vessel (white arrow). Magnification, ×20.

The authors thank Gerrard Lutty for help with the manuscript and Rob A. Elton for statistical assistance.