Effects of Oxygen and bFGF on the Vulnerability of Photoreceptors to Light Damage

Abstract

purpose. To test whether tissue oxygen levels affect the vulnerability of photoreceptors to damage by bright continuous light (BCL).

methods. Albino rats were raised in standard conditions of cyclic light (12-hour light, 12-hour darkness) with the light level at 5 to 10 lux or 40 to 65 lux. They were then exposed to BCL (1000–1400 lux), either continuously for 48 hours or for the day or night components of the 48-hour period. During BCL, some rats were kept in room air (normoxia, 21% oxygen), some in hypoxia (10%), and some in hyperoxia (70%). Their retinas were examined for cell death, for the expression of basic fibroblast growth factor (bFGF), and for response to light (electroretinogram, ERG).

results. The death of retinal cells induced by BCL was confined to photoreceptors. Within the retina, the severity of death was inversely related to the level of bFGF immunolabeling in the somas of the outer nuclear layer (ONL) before exposure. The death of photoreceptors was accompanied by an upregulation of bFGF protein levels in the ONL and by a decline in the ERG. Both hypoxia and hyperoxia during BCL reduced the photoreceptor death, bFGF upregulation, and ERG decline caused by BCL. The protective effects of hyperoxia and hypoxia were evident during both the day and night halves of the daily cycle. Hypoxia or hyperoxia alone did not upregulate bFGF or ciliary neurotrophic factor (CNTF) expression in the retina.

conclusions. Photoreceptors are protected from light damage by hypoxia and hyperoxia during exposure. The protection provided by oxygen levels operates during both day and night. The protection is not mediated by an upregulation of bFGF or CNTF.

Since the phenomenon of damage caused to the retina by bright continuous light (BCL) was first described, ^{1 2} many insights have been gained into the mechanisms of damage. There is evidence that the circadian rhythm of the retina is a major factor determining the severity of light damage ^{3 4} and that the circadian hormone melatonin and its receptors are of importance in mediating light damage. ^{5 6} Further, the retina, when stressed by light, upregulates oxygen radical scavengers ^{7 8 9} and can be protected from damage by an antioxidant applied exogenously, ⁹ suggesting that oxygen toxicity plays a role. The light-stressed retina also upregulates its expression of trophic growth factors, particularly basic fibroblast growth factor (bFGF) and ciliary neurotrophic factor (CNTF), ^{10 11} which are protective against a range of stresses. 

The mechanisms by which BCL damages photoreceptors and induces death are not more exactly known, however. In the present study we tested the influence of tissue oxygen levels during light exposure on the resultant damage. When evidence emerged that oxygen levels (both high and low) during BCL had a significant protective effect, the study was extended to assess whether these protective effects were mediated by bFGF and whether they were exerted during day or night halves of the day–night cycle. 

Methods

All procedures were in accord with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Light Management

Albino Sprague–Dawley rats were raised in cyclic light (12-hour light, 12-hour dark) with the light level either 5 to 10 lux or 40 to 65 lux. At 60 to 70 days of age, the rats, housed individually in standard rat boxes, were placed in a clear plexiglas chamber, and each rat was exposed to BCL from a fluorescent light placed above the chamber. The fluorescent light was a conventional cool, white light source, circular in shape. Measured as incident light by a photometer held at the floor of the cage, the brightness of the illumination was 1000 to 1400 lux. Most animals were exposed to BCL for 48 hours beginning immediately after the 12-hour period in darkness. During the 48-hour period of exposure to BCL, the rats were subjected to normoxia (room air, 21% oxygen), hypoxia (10% oxygen), or hyperoxia (70% oxygen). In a subset of experiments the rats were exposed to BCL and one of the same oxygen conditions for either the day or night halves of a 48-hour period. Food and water were available to the animals from containers kept on the floor of the cage. The top of the cage was kept free of containers, to minimize shading. The animals showed some initial tendency to hide from the bright light but adapted to the light within the first 12 hours, resuming feeding and grooming behavior. Persistent eyelid closure was not observed. The structure and visual function of the retina were examined at the end of the 48-hour period of exposure. Cell death and bFGF expression in the retina were examined in six experiments, each comprising three animals (one exposed in normoxia, one in hyperoxia, one in hypoxia). 

Oxygen Management

The level of oxygen in the plexiglas chambers was controlled by a feedback device (Oxycycler; Reming Bioinstruments, Redfield, NY), as described previously. ^{12 13 14 15} With the BCL source on, temperature in the chambers was 26°C to 27°C, 1°C to 2°C higher than the ambient temperature of the laboratory. Humidity in the chamber did not exceed 23%. 

Blood Gas Measurements

Arterial Po ₂ and Pco ₂ were measured on samples of blood drawn from rats while exposed to oxygen levels between 10% and 70%. With animals under surgical anesthesia, a cannula was placed in one carotid artery, externalized, and capped to allow a blood sample to be withdrawn from the conscious, freely moving animal. After 2 days’ recovery each rat was placed in an oxygen-controlled chamber. After a minimum of 30 minutes, a sample of 0.5 ml of blood was withdrawn in two stages. The first 0.25 ml was stored in a sterile syringe, and a further 0.25 ml was removed for testing. The first 0.25 ml was mixed with a similar volume of heparinized saline and returned to the circulation, through the carotid cannula. The second sample was analyzed for pH, Po ₂ and Pco ₂ using a blood gas analyzer (OPT1; AVL). The oxygen level was then reset and a further sample taken at a minimum 2-hour interval. A maximum of six samples was withdrawn in one day. 

ERG Recording

For electroretinogram (ERG) recording, rats were anesthetized with urethane (1.25 g/kg, intraperitoneally) or ketamine (100 mg/kg) and xylazine (12 mg/kg), intramuscularly. They were placed in a conventional head holder and kept warm with a feedback-controlled electric blanket. The pupil was dilated with topical 0.5% tropicamide and the cornea protected with a tear substitute gel. The ERG was recorded between a platinum loop touching the cornea and the pinna, from left and right eyes simultaneously. The frequency range of the amplifier was 0.3 to 200 Hz, and a 60-cycle notch filter was used. 

In practice, the anesthetic was administered a few minutes before the end of the 48-hour period of exposure to BCL and oxygen, and the animal was prepared for recording in room air and normal room light and placed in the recording apparatus. After 1 hour of dark adaptation, the ERG elicited by a bright flash, presented at intervals of 60 seconds or longer, was recorded. The stimulus was a flash unit placed 25 cm from the eyes. The power of the flash, measured through a 3-mm aperture at 25 cm, was 8 μW. The amplitude was recorded as the signal range in a specified time frame (Fig. 9) . When an a-wave was present, this measure showed the voltage difference between the negative (a-wave) and the positive (b-wave) peaks. When the a-wave was absent, this measure showed the b-wave amplitude. Stimulus intensity could be attenuated in 0.5-log-unit steps with neutral-density filters. 

Tissue Preparation

Morphologic observations were made on eyes that were not studied electroretinographically, in case the light and oxygen conditions during the period of recording changed the effects induced by BCL. Protocols for fixation, cryoprotection, and sectioning of the eyes were as previously described. ¹³ Sections were labeled for cell death using the TdT-dUTP terminal nick-end labeling (TUNEL) technique, ¹⁶ according to previously described protocols, ¹² and for bFGF and glial fibrillary acidic protein (GFAP) or vimentin or CNTF. The immunocytochemistry protocols have been described previously, ¹³ except for CNTF. We used a rabbit polyclonal antibody to rat CNTF (Chemicon, Temecula, CA), made up in 0.3% Triton/1% bovine serum albumin (BSA) solution at a dilution of 1:200 and followed the protocols previously described for anti-bFGF labeling. 

Some sections were also labeled with a DNA-specific fluorescent dye, using a green fluorescent dye (Syto; Molecular Probes, Eugene, OR) or bisbenzamide (Calbiochem, La Jolla, CA). Sections were pretreated with 70% ethanol (20 minutes), washed with phosphate-buffered saline (PBS), and incubated in the dye solution. Incubation time for the fluorescent dye (Syto 12 diluted 1:1000 in PBS; Molecular Probes) was 30 seconds and for bisbenzamide (1:2000) was 10 minutes. 

Tissue Assessment

TUNEL⁺ profiles were counted and analyzed as previously described. ¹³ For each retina, four sections were counted, and the counts were averaged. In regions of retina where TUNEL was maximal, the labeling was so dense that individual profiles could not be distinguished. The count was then recorded as 500 per 400μ m, a value slightly above the highest practical count. 

Sample areas of the retinas were imaged in a confocal microscope equipped with both argon-krypton and UV lasers. When we sought to compare the density of bFGF labeling between sections from different retina, the eyes were processed (blocked, cut, labeled, examined by confocal microscopy) in the same sessions and, during confocal microscopy, photomultiplier settings were held constant. Signal intensity was then measured with NIH Image software, using the Analysis tool (provided in the public domain by the National Institutes of Health, Bethesda, MD). 

Results

Arterial Gas Levels in Hyperoxia and Hypoxia

PAo ₂ varied approximately linearly over a wide range of levels of oxygen inhaled (Fig. 1) . Pco ₂ varied within a narrow range, decreasing when oxygen levels were below 21%. Correspondingly, arterial pH increased when oxygen levels were below 21%, and in one animal pH decreased at high oxygen levels. The data in Figure 1 confirm previous observations ¹⁷ for inhaled oxygen levels higher than 21%. The data for oxygen levels lower than 21% are novel. Because oxygen reaches the outer retina by diffusion from the choriocapillaris and Po ₂ levels in the choriocapillaris are close to arterial levels, ¹⁸ these measures provide an estimate of maximal oxygen levels available to outer retina. 

Adult Rat Retina before Light Challenge

In the adult Sprague–Dawley rat retina, very few cells were dying; only the occasional cell was TUNEL⁺ (red in Fig. 2A ). Higher densities of TUNEL⁺ profiles (as in Fig. 2B ) resulted from damaging stimuli, in this case 24 hours of BCL. Astrocytes (identified by their GFAP content, red in Fig. 2C ) were largely confined to the inner surface. bFGF (green in Figs. 2C 2D 2E 2F 2G 2H 2I ) was prominent in somas in the inner nuclear layer (INL; i in Fig. 2C ) and in astrocyte nuclei (e.g., top of Fig. 2C ). At the peripheral edge of the retina (Fig. 2D , lower arrow) bFGF was prominent in the outer nuclear layer (ONL; o in Fig. 2 ) and, toward the edge (between the arrows in Fig. 2D ), GFAP expression (red) in Müller cells appeared upregulated. The presence of GFAP in Müller cell processes is an index of environmental, ^{19 20 21} genetic, ^{19 22} and mechanical ^{11 23 24 25} stress. The co-upregulation of bFGF in the ONL and of GFAP in Müller cells was normally apparent only at the edge of the retina but could be elicited in midperipheral retina by exposure to BCL (Fig. 2E) . In favorable preparations of unstressed retina, bFGF protein could be seen in the end feet of Müller cells (Fig. 2F , vertical arrow) as well as astrocyte nuclei (oblique arrow). The Müller cell processes, labeled for vimentin (red) in Figures 2F 2G and 2H , could be followed to bFGF-laden somas in the INL. At higher power (Fig. 2H) each vimentin bundle could be seen to skirt or split around a bFGF-laden profile, suggesting that the bFGF profile was the nucleus of a Müller cell. Around retinal vessels (Fig. 2I) , the astrocytic glia limitans was GFAP⁺ (red). bFGF was prominent in astrocyte nuclei (Fig. 2I , arrows), and a blush of green labeling inside the glia limitans indicated the presence of bFGF protein in the endothelial cells of the vessel wall. 

Effects of Hypoxia and Hyperoxia on Photoreceptor Death

TUNEL labeling induced in the retina by BCL was specific to the ONL and therefore to photoreceptors (Fig. 3) . TUNEL was observed in the INL only in areas where labeling of the ONL was particularly intense (Fig. 3B) . This INL labeling was almost certainly a secondary effect due to uptake of TUNEL-labeled DNA of photoreceptors by Müller cells. ²⁶  

In rats raised in 40 to 65 lux and exposed to BCL in normoxia, TUNEL labeling was most prominent in superior midperipheral retina (Fig. 3B) , less marked in the inferior midperiphery (Fig. 3C) , and least visible at the edges of the retina (Figs. 3A 3D) . Hyperoxia during BCL reduced the level of photoreceptor death caused by BCL, most obviously in the superior midperiphery (compare Figs. 3B and 3F ) but also in the inferior midperiphery (compare Figs. 3C and 3G ). The protective effect of hyperoxia was least obvious at the retinal edges, where levels of death in normoxia were lowest (compare Figs. 3A and 3E 3D and 3H ). Hypoxia during BCL was also protective. Again, the protection was most evident in the superior midperiphery (compare Figs. 3B and 3J ) and least evident at the edge of the retina (compare Figs. 3A and 3I , 3D and 3L ). 

In rats raised in 5 to 10 lux, the region of retina in which photoreceptor degeneration was extensive was considerably wider, extending from superior retina into inferior retina (data not shown). Nevertheless, both hyperoxia and hypoxia reduced the level of death (below). 

TUNEL labeling was quantified in six triple experiments, in each of which one animal was exposed to BCL in normoxia, one in hypoxia, and one in hyperoxia. Three of the triple experiments involved animals raised in dim (5–10 lux) cyclic light; three were raised in brighter (40–65 lux) cyclic light. Pooling counts from all areas of retina and both conditioning regimens (Fig. 4) , the frequency of TUNEL⁺ profiles was lower in both the 70% and 10% groups, and the differences between oxygen-treated and control groups were significant (P ≪ 0.001 on a t-test). 

When dim- and bright-light–raised groups were separated (Fig. 4B) , it was evident that cell death was higher in the dim-raised group in animals kept in room air during exposure to BCL (confirming the findings of Penn and Anderson ²⁷ ). Further, both hyperoxia and hypoxia reduced the cell death induced by BCL in both dim- and bright-reared groups, but their protective effect was weaker for the dim-reared group. 

The effects of oxygen on cell death induced by BCL in superior and inferior retina are separated in Figs. 4C and 4D . In bright-reared animals, oxygen levels caused significant reduction in the death induced by BCL in both superior and inferior retina (Fig. 4C) . In the dim-reared group (Fig. 4D) the protective effects were weaker and, in the present data, reached significance in the superior retina only in hyperoxia and in the inferior retina only in hypoxia. 

Gradient in bFGF Expression before Exposure to BCL

In animals raised in both dim and bright conditions, bFGF protein was prominent in photoreceptor somas at the edges of the retina, where, for a short length of the section, bFGF appeared to fill the cytoplasm of all ONL somas (Figs. 5A 5F ) and GFAP was upregulated in Müller cells. More centrally, Müller cell processes did not label with GFAP, and bFGF was less prominent in the ONL. In animals raised in dim conditions, bFGF was not detected in photoreceptors in the central or midperipheral retina. In animals raised in 40 to 65 lux, bFGF was present in photoreceptor somas in midperipheral retina, more prominently in inferior than in superior retina. In superior retina 1 to 2 mm above the optic disc, bFGF labeling in the ONL was faint and formed a network, surrounding the somas (Fig. 5B) . Nearer the optic disc (Fig. 5C) the bFGF labeling in the ONL was stronger, showing the network pattern more clearly. Several millimeters inferior to the optic disc (Figs. 5D 5E) bFGF label was also present in the somas of ONL cells. 

Comparing Figures 3 and 5 suggests that the level of death induced by BCL was inversely proportional to the level of bFGF in ONL somas before exposure. Thus, death was least at the edges of the retina (Figs. 3A 3D) , where bFGF was highly concentrated in ONL somas (Figs. 5A 5F) , and most severe in retina 1 to 2 mm superior to the optic disc (Fig. 3B) , where bFGF was not detected in ONL somas (Fig. 5B) . 

Role of Hypoxia Mediated by Trophic Factors

If oxygen (high or low) protects by upregulating trophic factors such as bFGF and CNTF, then it should be possible to demonstrate regulation of these factors by oxygen alone. To test this, we examined the retinas of rats raised in dim conditions and exposed to 70%, 21%, or 10% oxygen levels for 7 days. In control retinas (kept in 21% oxygen, i.e., room air) the patterns of bFGF and CNTF seen in midperipheral retina and at the retinal edge (Figs. 6C 6D ) confirmed previous descriptions. ^{14 28} bFGF was prominent in somas of Müller cells in the INL and in astrocyte nuclei at the inner surface, but was not prominent in photoreceptor somas except at the retinal edge (Fig. 6D) . CNTF was detected in astrocyte and Müller cell processes, including those in the outer limiting membrane (OLM). At the edge of the retina, CNTF was upregulated but remained within macroglial (astrocyte and Müller cell) nuclei and processes. Quantitatively, the accumulation of bFGF in ONL somas at the edge of the retina was very prominent (Fig. 6J) as was the upregulation of CNTF in the OLM (Fig. 6L) . Retinal layers were considerably thinner at the edge of the retina than in the midperiphery. Assessed by inspection (compare Figs. 6A 6C and 6E ) or quantitatively (Figs. 6K 6M) , exposure to 10% or 70% oxygen for 7 days did not produce significant upregulation of either bFGF or CNTF. 

Conversely, if the protective effects of oxygen are mediated by their upregulation of trophic factors, then in areas where oxygen reduces the photoreceptor death caused by BCL, there should be evidence of an upregulation of bFGF and, where the protection is least, the expression of bFGF should be least. In fact, the converse was the case. For example, in the superior midperiphery (Fig. 7A ) of the retina exposed to BCL without oxygen protection, photoreceptor death was greatest, and bFGF levels were spectacularly high. Conversely, where oxygen protected the superior midperipheral retina, bFGF levels were much lower (Figs. 7D 7G) . A similar but less spectacular trend was apparent in the inferior midperiphery (Figs. 7B 7E 7H) . BCL had the least effect on bFGF protein levels at the edge of the retina, where levels before BCL exposure were high, and the death of photoreceptors induced by BCL was minimal (Fig. 7C) . Shown quantitatively (Figs. 7J 7K) , the level of bFGF after BCL was markedly elevated in the superior midperiphery of the normoxic retina, but not in the retinas exposed simultaneously to oxygen. Similarly, the protective effects of oxygen on the inferior midperiphery of the retina were not associated with increased bFGF levels (Fig. 7K) . In the normoxic retina, the ONL was abnormally thick, a result of the swelling and fragmentation of the disintegrating nuclei. 

Effects of Hypoxia and Hyperoxia on BCL-Induced bFGF Upregulation

In retinas exposed to BCL in hyperoxia and hypoxia, the upregulation of bFGF induced by BCL was less than in retinas exposed in normoxia. This was apparent in both the superior midperipheral retina (compare Figs. 7A with 7D and 7G) and inferior midperipheral retina (compare Figs. 7B 7E and 7H ). At the edges of the retina, where bFGF levels in the ONL were high before exposure to BCL, and the death induced by BCL and the reductions in cell death associated with oxygen were least, oxygen had little effect on the level of bFGF in the ONL (Figs. 7F 7I) . 

Day–Night Experiments

To test whether the protective effects of hyperoxia and hypoxia were exerted during day or night, rats raised in 5- to 10-lux cyclic light were exposed to BCL and to hyperoxia or hypoxia for only the day halves of the 48-hour test period or for only the night halves of the period. The retinas of all animals were examined at the end of the 48-hour period. In all four variants of these protocols (hyperoxia and BCL at night, hyperoxia and BCL by day, hypoxia and BCL at night, hypoxia and BCL by day), photoreceptor death assessed by the TUNEL technique was lower than in the normoxic controls (Fig. 8A ) and the effects of hyperoxia and hypoxia were apparent in all regions of the retina, during both night and day (Figs. 8B 8C) . 

Hypoxia and Hyperoxia during BCL Limited the Decline of the ERG

In retinas exposed to BCL in normoxia, the amplitude of the dark-adapted flash-evoked ERG was lower than in normal retinas (Fig. 9) . The a-wave was inconspicuous, the peak of the b-wave was delayed from approximately 80 msec in the normal retina to 100 msec or more and the duration of the b-wave was increased. ERGs recorded after BCL in hypoxia and hyperoxia were abnormal in the same ways (Figs. 9B 9C) . 

The ERG was recorded in seven animals after exposure to BCL in normoxia, in six animals after exposure in hypoxia, and in five animals after exposure in hyperoxia. Variability of amplitude was evident in all BCL-exposed animals, as reported by previous investigators. ⁶ Considerable variability remained after we standardized the light history of the animals, the period of dark adaptation (minimum, 1 hour) before recording and the interflash interval (minimum, 60 seconds) and was not markedly reduced when ketamine-xylazine anesthesia was used instead of urethane. ERG amplitude ranged from 12 to 407 μV among animals exposed to BCL in normoxia, from 298 to 626 μV among the animals exposed in hypoxia, and from 340 to 465 μV among the animals exposed in hyperoxia. 

For the following analyses, we used the responses from the left eye, unless the left eye traces were unavailable. For each animal we selected the last six responses recorded in stable conditions. Traces were pooled into three groups, yielding mean ERG amplitudes of 321 μV for animals exposed in normoxia, 417 μV for animals exposed in hypoxia, and 411 μV for animals exposed in hyperoxia. When a t-test was applied to these groups of traces, the probability that the hyperoxic and normoxic amplitudes could be obtained from the same population of responses was low (P < 0.01); for hypoxic and normoxic traces, P was also low (< 0.01). However, an autocorrelation analysis indicated significant serial dependence between the readings obtained from one animal, so that a t-test, which assumes independence of responses, was inappropriate. The numbers of animals ^{5 6 7} were too small for a t-test to be applicable to the means from each animal. We therefore used the nonparametric Wilcoxon two-sample test. This yielded P = 0.037 for the hypothesis that the hyperoxic and normoxic traces came from the same population of traces. For the hypoxic and normoxic values, P = 0.069. 

Finally, we examined the effect of oxygen levels during BCL exposure on ERGs elicited by submaximal stimuli (Fig. 9D) . For all three groups (hypoxic, normoxic, hyperoxic), ERG amplitude declined as stimulus intensity was attenuated. Over the 3.0-log-unit range tested, response amplitudes from animals exposed to BCL in normoxia were smaller than those from animals exposed in hypoxia and hyperoxia. 

Discussion

The present results provide evidence that BCL induced death specifically in photoreceptors; that retinas of rats raised in physiological conditions could contain significant gradients of bFGF protein expression in photoreceptors; that the death of photoreceptors induced by BCL occurred in gradients inversely related to those prior bFGF gradients; that hyperoxia and hypoxia during BCL reduced the photoreceptor death and ERG decline induced by BCL; that the protection provided by hyperoxia and hypoxia acted during both day and night; and that the protection provided by hyperoxia and hypoxia was not mediated by bFGF or CNTF. 

bFGF in Photoreceptors: Regulated by Stress, a Factor in Survival

Accumulation of bFGF protein in photoreceptors is prominent when the retina is damaged—for example, by laser energy ^{28 29} —or is affected by damage to the optic nerve. ³⁰ Upregulation of bFGF mRNA expression in rat photoreceptors has been reported after light damage ³¹ and gene-induced degeneration. ^{31 32} bFGF is prominent in the ONL of degenerating human retina ³³ and is upregulated in the ONL of the degenerating mouse retina. ³² Other investigators ^{34 35} have reported specific immunolabeling of bFGF in several classes of retinal cells, but not of photoreceptors. The present observations were made using the same antibody as Xiao et al. ²⁸ and confirm their observation of strong bFGF immunoreactivity in photoreceptors at the retinal edge. 

Present results and studies continuing in this laboratory ¹⁴ provide evidence that light stress that induces photoreceptor death also upregulates the level of bFGF protein in the ONL and suggest that lesser but significant levels of bFGF accumulate in photoreceptors conditioned to relatively bright daily light. Penn and Anderson ³⁶ reviewed extensive evidence that the light history of the retina determines the severity of damage induced by BCL. The relative resistance of inferior retina in our experiments could be explained in these terms. It is possible that inferior retina experienced higher levels of light during conditioning, because the lights in the rooms in which our rats were raised were located in the ceiling. 

Earlier studies have provided evidence that stimuli that damage photoreceptors, whether by heat, ³⁷ needlestick injury, ³⁸ laser coagulation injury, ^{28 39} or damaging light, ¹⁰ make photoreceptors that survive the damage resistant to subsequent challenge. For all except heat, for which the point was not tested, the protection was shown to be associated with an upregulation of bFGF protein or mRNA in the retina. Further, exogenously applied bFGF protects photoreceptors against light damage, ⁴⁰ and a stimulus that upregulates bFGF in ONL somas without killing them ^{30 41} also increases their resistance to damaging stimuli. ⁴² Present observations suggest that, within the same retina, the level of damage caused by BCL varied with the level of bFGF protein in photoreceptor somas before exposure to BCL and add the point ¹⁴ that the level of bFGF protein in the ONL appeared to be regulated significantly by physiological ranges of light experience, as well as by trauma. 

The upregulation of bFGF at the edge of the retina is also strongly correlated with protection of photoreceptors (Fig. 3) . The strong expression of bFGF in photoreceptors at the edges of the retina (Figs. 2D 5A 5F) colocalizes with an upregulation of GFAP in Müller cells, suggesting that it is caused by (still unidentified) stress. 

Variability in ERG Effects

Variability in the damage produced in the rat retina by exposure to BCL has been reported by previous investigators and was evident in both our morphologic and ERG data. The variability was greater in the ERG data, even after we had controlled light history, anesthesia, interflash intervals, and duration of dark adaptation, and the discrepancy deserves comment. The morphologic data indicated the photoreceptors with DNA that had begun to fragment in a particular condition, signaling apoptotic death. The ERG measured the summed signal elicited from photoreceptors by a flash. It is possible that some photoreceptors survived BCL (were TUNEL⁻) but with damaged outer segments that could produce only weak responses; and it is possible that photoreceptors that had begun to die, produced a measurable response to a light flash. 

Mechanisms of Hypoxia- and Hyperoxia-Induced Protection

The protection provided by hypoxia to light-challenged photoreceptors can be explained in terms developed previously. Light saturates rods, reducing their metabolism and therefore their oxygen consumption, oxygen levels in the outer nuclear layer increase, ⁴³ the frequency of oxygen radicals presumably increases, and these radicals may induce apoptosis in photoreceptors. Antioxidants have been reported to be protective against light damage, ⁹ and endogenous radical scavenger mechanisms are upregulated in retinas that are resistant to light damage. ⁸ In his analysis of the effect of hypoxia on the retina Steinberg ⁴⁴ concluded that light protects the retina from the damaging effects of hypoxia. The present results suggest that hypoxia protects the retina from the damaging effects of light. 

This line of argument predicts, however, that hyperoxia exacerbates light damage and, in accord with that prediction, Ruffolo et al. ⁴⁵ reported that raised arterial Po ₂ decreases the threshold for light-induced damage. In the present experiments, however, hyperoxia was protective against light damage, whether assessed by TUNEL assay and morphologic damage, or by the ERG, suggesting that quite different mechanisms of damage are involved. Ruffolo et al. studied the monkey and used extremely bright light, from a xenon arc source, focused on a small patch of retina. The effect of this light was to produce a visible lesion within 100 seconds, with the principal damage occurring in the pigment epithelium. Their estimate of the power of their light indicates that it was equivalent to 7 to 25 × 10⁵ lux, 2 orders of magnitude brighter than that used in the present experiments (1.0 to 1.4 × 10³ lux). These latter levels of intensity were at the high end of the physiological range, were damaging only if left on for periods of hours to days, and caused damage principally in photoreceptors. 

The protective effect of hyperoxia is thus difficult to explain in the same terms as for hypoxia, and we investigated a possible role of circadian mechanisms. Recent studies have described a day–night change in neurohormones released in the retina, with melatonin released by photoreceptors at night and dopamine released by amacrine cells in response to daylight. Further, melatonin inhibits dopamine release, and vice versa, suggesting that the neurohumoral environment of the retina switches between day and night (reviewed in References ^{46 47 48 49} ) and melatonin levels and melatonin receptor function have been shown to modulate the retina’s vulnerability to light damage. ^{5 6 50} It seemed possible that hyperoxia may protect during one part of the circadian cycle, hypoxia during another. The present results (Fig. 8) confirm the ability of both hyperoxia and hypoxia to protect photoreceptors, but do not support the idea that hypoxia- and hyperoxia-induced protective mechanisms are separated within the day–night cycle. 

Finally, we tested whether hyperoxia and hypoxia both activate a common protective mechanism, by upregulating trophic factors such as bFGF and CNTF (Figs. 6 7) , as in the preconditioning paradigm of Liu et al. ¹⁰ We could not, however, demonstrate that either hyperoxia or hypoxia upregulates bFGF levels over the exposure periods used and, conversely, we could not demonstrate that, where oxygen (high or low) provides protection, bFGF is upregulated. Indeed, the regulation of bFGF by BCL in the presence of oxygen seemed the opposite of that predicted if either hypoxia and hyperoxia protects by upregulating the factor. 

It seems reasonable to conclude that neither hypoxia nor hyperoxia protects by upregulating mechanisms that are activated when the retina is preconditioned—for example, by exposure to a potentially damaging stimulus, such as bright light. ¹⁰ The protective effects of oxygen may then act upstream from the mechanisms that activate both the death of photoreceptors and the upregulation of protective factors. This last suggestion is speculative, however, and for the present the mechanisms of the protective actions of hypoxia and particularly of hyperoxia remain unresolved. 

Dim Versus Bright Rearing, Day Versus Night

The protective effects of hyperoxia and hypoxia were apparent in both dim-reared and bright-reared groups (Fig. 4) and during both day and night (Fig. 8) . The protective effects were weaker in dim-reared than in bright-reared animals. This may be because the dim-reared retinas were more vulnerable to damage by BCL, and it is possible that a clearer rescue effect could be demonstrated with a weaker BCL challenge. 

The protective effects of hyperoxia and hypoxia were clear in both day and night data. There was a consistent trend in the day–night data (Fig. 8) for photoreceptor death induced by BCL to be greater during day than during night. This seems at odds with a previous report ⁴ that the rat retina is more vulnerable to light damage during the night, but the design of present experiments was not appropriate to test the specific period of high vulnerability, in the early morning hours, described in that study. 

 Supported by Retina Australia, the National Health and Medical Research Council of Australia, and the Medical Foundation of the University of Sydney.

 Submitted for publication April 19, 2000; revised July 10 and October 17, 2000; accepted November 8, 2000.

 Commercial relationships policy: N.

 

Figure 1.

 

Blood gas measurements in normal adult Sprague–Dawley rats, as a function of the proportion of oxygen in the air inhaled. (A) Arterial oxygen (Pao ₂) and carbon dioxide (Paco ₂) for three experimental animals. (B) Arterial pH for the same animals, assessed from the same blood samples.

Figure 2.

 

TUNEL and bFGF labeling of rat retina. The blue label is bisbenzamide staining of normal DNA. The green labeling is the antibody to bFGF. (A) TUNEL-labeled cells were rare in the normal adult retina. This example (red) is in the ONL (o). (B) The frequency of TUNEL⁺ cells was raised by exposure to BCL, in this case for 24 hours. (C) Astrocytes (labeled red with an antibody to GFAP) were largely restricted to the inner surface of the retina. Occasionally, an astrocyte process (arrow) followed a blood vessel into the inner plexiform layer (IPL). bFGF was prominent in astrocyte nuclei (example at top) and in nuclei of the inner nuclear layer. (D) The peripheral 1 mm of normal retina. At the very edge (lower arrow) bFGF was very prominent in the ONL. GFAP-labeling (red) was prominent in Müller processes (between the arrows). (E) In light-damaged retina bFGF was prominent in photoreceptor somas in the ONL, and GFAP was upregulated in the radial processes of Müller cells. (F) Vimentin (red) labeling shows Müller processes running across the IPL to bFGF-laden profiles of the INL. The red signal is removed over a small patch (upper left; vertical arrow) to show bFGF (green) in the inner end feet of Müller cells. (G) In some preparations, bFGF-labeled granules (arrows) could be seen distributed along the processes of vimentin-labeled (red) Müller cells in the IPL. (H) Vimentin (red)-labeled Müller cell processes become closely apposed to bFGF-laden profiles in the INL (i). (I) Vessels at the inner surface of the retina were ringed by astrocytes processes, labeled red with the antibody to GFAP. This is the vessel’s glia limitans. BFGF-labeling was strong in astrocyte nuclei (arrows) and colocalized with bisbenzamide labeling of endothelial cell nuclei inside the glia limitans. Scale bars, (A, B, and C) 50 μm; (D) 100 μm; (F through I) 20 μm.

Figure 3.

 

Distribution of fragmenting DNA identified by TUNEL-labeling (red) at the end of exposure of the retina to BCL for 48 hours. The blue label shows nonfragmenting DNA labeled with bisbenzamide. Regions of retina from animal exposed to BCL in (A through D) normoxia (room air, 21% oxygen), (E through H) hyperoxia (70% oxygen), (I through L) and hypoxia (10% oxygen). (F) g, ganglion cell layer; i, INL; o, ONL.

Figure 4.

 

The influence of oxygen on cell death. (A) Pooled data from six triple experiments (one animal exposed to BCL in 21% oxygen, one in 70%, one in 10%) showed that photoreceptor death was significantly less in the 70% and 10% groups. (B) When data for animals raised in 40 to 65 lux (bright-reared) were separated from those of animals raised in 5 to 10 lux (dim-reared), death rates were higher in the dim-reared animals, and the protective effects of oxygen were weaker. (C, D) Separating superior from inferior retina, the protective effects of hyperoxia and hypoxia were both significant in the bright-reared animals. In dim-reared animals (D) hypoxia was protective in inferior, but not in superior, retina, whereas the reduction in cell death produced by hyperoxia was significant only in superior retina.

Figure 5.

 

Distribution of bFGF protein in the retina before exposure to BCL. Green label: binding of an antibody to bFGF; red: binding of an antibody to GFAP. (A) The superior edge of the retina. Small arrow: the edge; large arrow: optic disc. (B, C) Retina superior to the optic disc. (C) is nearer the optic disc than (B) and (D). (E) Retina inferior to the optic disc. (D) is nearer to the optic disc than (E). (F) Inferior edge of the retina. Small arrow: the edge; large arrow: optic disc. At the edge of the retina (A, F) bFGF was strongly upregulated in the somas of photoreceptors (green), and GFAP was upregulated in the processes of Müller cells (red).

Figure 6.

 

Effects of oxygen on the expression of bFGF and CNTF protein. Green: BFGF immunolabeling; red: CNTF. Superior midperipheral and edge of a rat retina (A, B) after 7 days in hyperoxia, (C, D) in normoxia (21%), (E, F) and after 7 days in hypoxia. (J, K) Levels of bFGF labeling across midperipheral and edge regions of a normoxic retina (J) and across midperipheral regions of retinas raised in normoxia, hyperoxia, and hypoxia (K). (L, M) Levels of CNTF labeling across midperipheral and edge regions of a normoxic retina (L) and across midperipheral regions of retinas raised in normoxia, hyperoxia, and hypoxia (M). ILM, inner limiting membrane; OLM, outer limiting membrane.

Figure 7.

 

Distribution of bFGF protein in the retina immediately after exposure to BCL. Each of (A) through (I) shows the bFGF protein label in a region of retina (left or top panel) and the cellular structure of the retina, demonstrated with a DNA label (right or lower panel). (A, B, and C) Retina exposed to BCL in normoxia (room air, 21% oxygen). (A) In the sensitive region in superior retina bFGF was observed in Müller cell somas in the INL (as in normal retina). In addition the ONL was extensively labeled . (B) The ONL was less strongly labeled in inferior midperipheral retina. (C) At the edge of the retina the ONL was strongly labeled for bFGF, as in the normal retina (Fig. 3) . (D, E, and F) Corresponding regions from a retina exposed to BCL in hyperoxia (70% oxygen). The upregulation of bFGF labeling in the ONL seen in normoxia (A, B) was not apparent. (G, H, and I) Corresponding regions from a retina exposed to BCL in hypoxia (10% oxygen). The upregulation of bFGF labeling in the ONL seen in normoxia (A, B) was not apparent. (J) bFGF immunolabeling intensity across the thickness of superior midperipheral retina, from rats exposed to BCL in three oxygen condition. (K) Same as (J), but for the inferior midperipheral retina.

Figure 8.

 

Effects of day and night exposure on oxygen-mediated protection of photoreceptors. (A) Both hyperoxia and hypoxia were associated with reduced photoreceptor death, for both day exposure and night exposure. The protective effects of oxygen were apparent in all areas of retina for night (B) and day (C). On a two-tailed t-test of the differences between the 70% and 21% counts, and between the 10% and 21% counts, P ≪ 0.001, for all comparisons.

Figure 9.

 

Representative ERG recordings from eyes exposed to BCL. Recordings were made after 1-hour dark adaptation in response to a full-intensity flash. Except for this delay, the recordings were made as soon as practical after the end of exposure to BCL. (A) Black traces: the two eyes of one animal in which the ERG was recorded after exposure to BCL in normoxia (room air, 21% oxygen). Gray trace: ERG from a normal animal not exposed to BCL, taken after an approximately 1-hour dark adaptation from room light (60–100 lux). (B) An animal exposed to BCL in hyperoxia (70% oxygen). (C) An animal exposed to BCL in hypoxia (10% oxygen). (D) Effect of oxygen on ERG amplitude over a stimulus range. When the flash intensity was attenuated with neutral density filters, ERG amplitudes decreased for all three groups. Over this range the response of those exposed in 21% oxygen was consistently lower than the response of those exposed in higher or lower oxygen levels. The numbers in columns above each data point are the number of animals tested for the mean value shown, with the hyperoxic n at top, and the normoxic n at bottom.

The authors thank Eugene Seneta for statistical advice.