Retinal Oxygen: Fundamental and Clinical Aspects

We reviewed research on retinal oxygen (O₂) distribution and use, focusing on O₂ microelectrode studies in animals with circulatory patterns similar to those of humans. The inner and outer halves of the retina are different domains in terms of O₂. Understanding their properties can suggest mechanisms of and therapies for retinal diseases. Inner retinal PO₂ averages about 20 mmHg. Effective O₂ autoregulation of the retinal circulation ensures that inner retinal PO₂ is relatively uninfluenced by systemic hypoxia and hyperoxia and increased intraocular pressure in healthy animals. Failures of the retinal circulation lead to tissue hypoxia that underlies the vasoproliferation in diabetic retinopathy and retinopathy of prematurity. Choroidal blood flow is not regulated metabolically, so systemic hypoxia and elevated intraocular pressure lead to decreases in choroidal PO₂ and photoreceptor O₂ consumption. The same lack of regulation allows choroidal PO₂ to increase dramatically during hyperoxia, offering the potential for O₂ to be used therapeutically in retinal vascular occlusive diseases and retinal detachment.

Oxygen (O₂) is essential for retinal function. As in other tissues, O₂ diffuses through the tissue passively from the circulation and is consumed in the oxidative production of adenosine triphosphate. However, several features make retinal oxygenation unique, and these have an impact on the pathogenesis and treatment of retinal diseases. Among the unique features influencing retinal oxygenation are the presence of a dual circulation, lack of metabolic regulation of the choroid, presence of O₂ regulation of the retinal circulation, and localization of mitochondria to photoreceptor inner segments. In most cases, it is useful to think of the retina as the following 2 domains: the avascular outer retina and the vascularized inner retina. The retina is also one of the most metabolically active tissues, consuming O₂ more rapidly than many other tissues, ¹ including the brain.^2,3

Because the demand for O₂ is high, and because it cannot be stored in the retina or other tissues, a continuous O₂ supply is essential.⁴ This review focuses on vascularized retinas (eg, of the rat, cat, pig, monkey, and human) in which O₂ is delivered by the retinal and choroidal circulations. In some species, retinal vasculature is nearly or completely absent (eg, the rabbit and guinea pig). These retinas rely on the choroid almost exclusively and the inner retina is strongly glycolytic.^5-7 A review focusing on avascular and partially vascularized retina has been presented elsewhere.⁸

Retinal O₂ tension (PO₂) and other variables that provide an index of PO₂ have been measured in several ways, which have recently been presented by Hogeboom van Buggenum et al.⁹ Hemoglobin saturation in retinal vessels has been measured by several groups.^10-12 Intravascular PO₂ has been measured with a phosphorescent dye, for which decay is inversely proportional to the PO₂.¹³ Berkowitz and coworkers^14,15 have pioneered the development of magnetic resonance imaging techniques that measure absolute PO₂ in the preretinal vitreous of animals and humans using the fluorine signal from perfluorocarbon droplets. They have also measured relative PO₂ by using the dependence of the hydrogen signal on O₂.¹⁶ However, only oxygen microelectrodes^17-20 have been able to map out O₂ gradients within the retina with high spatial and temporal resolution. We will focus on these measurements in the first part of this review. In some of this work, a second barrel of the electrode has been used to record the local electroretinogram (ERG) and the transepithelial potential across the retinal pigment epithelium, which allows verification of the position of the intraretinal electrode.^21,22 The invasive nature of the microelectrode is the major disadvantage of this technique, because the electrode may damage the tissue during measurement. Hence, this technique has limited clinical application. To date, microelectrodes have only been used in humans to measure O₂ in the vitreous during intraocular surgery.^23,24 Some authors have argued that O₂ consumption by the microelectrode itself is another disadvantage.¹² However, small recessed cathode microelectrodes^17,20 do not significantly disturb the local O₂ environment.²⁵ Microelectrode O₂ consumption (QO₂) is about the same as that of a single small cell. An electrode producing a current of 2 pA, which is typical in our work, uses 5.2 × 10^{− 16} mol of O₂ per second. This is the same as the O₂ use of a 6.2-µm cube of tissue in a region with a modest QO₂ of 3 mL O₂ per 100 g of tissue per minute.

Oxygen recordings of the type presented herein are obtained as shown in Figure 1. The microelectrode is advanced with a microdrive through the retina in discrete (eg, 3-µm) steps until it reaches the choroid. The electrode is then withdrawn continuously, usually at a constant speed of 2 µm/s in our work. The map of PO₂ across the retina, an O₂ profile, can be recorded during microelectrode penetration or withdrawal, and these give similar results. Data obtained during withdrawals are preferable because one can be sure that the tissue and vasculature are not being compressed, and because a larger number of data points can be recorded than during penetration. In other experiments, the microelectrode has been left stationary at a point in the retina or vitreous to study responses to light, ^22,26 effects of altered inspired gases, ^27-30 and spontaneous fluctuations in retinal PO₂.³¹

In vivo O₂ profiles have been recorded from pigs, ^21,32 cats, ^22,33 rats, ³⁴ and monkeys.³⁵ These studies have shown similar O₂ profiles everywhere except in the monkey fovea. Under light-adapted conditions (ie, enough light to saturate rod responses), the PO₂ falls steeply between the choriocapillaris(approximately 60 mmHg) and the photoreceptor inner segments and subsequently decreases more gradually through the outer retina, as shown in Figure 2. The direction of this gradient indicates that all of the O₂ used by the photoreceptors comes from the choroid under light-adapted conditions. In darkness, the PO₂ in the outer retina is considerably lower, ^22,35 reaching 0 mmHg at the proximal side of the inner segments. Sometimes much of the outer nuclear layer in the cat has a PO₂ of 0 mmHg (Figure 2). This low value leads to a reversal of the gradient of O₂ proximal to the inner segment, so that in darkness, O₂ also diffuses from the retinal circulation to the photoreceptors.

The steep drop of PO₂ from the choroid reflects the high rate of O₂ use by the photoreceptors.^36,37 The high choroidal PO₂ is maintained by the unusually high flow rate in the choroidal circulation, ^38-40 which is necessary for providing enough O₂ to the retina.⁴¹

One or more O₂ peaks appear in the inner retina (Figure 2). These indicate O₂ sources and, therefore, the presence of nearby retinal vessels. The preretinal PO₂ in the vitreous is normally a good indicator of inner retinal PO₂, ^42,43 because the vitreous consumes very little O₂, ⁴⁴ and the inner retina is the source of vitreal O₂.^43,44 The average PO₂ in the inner retina of the cat during dark adaptation was 18.5 mmHg, ⁴⁵ which is in good agreement with the preretinal PO₂ in cats of 18.9, ²⁸ 20 to 30, ⁴⁴ 15 to 20, ⁴² 20.2, ³³ and 19 mmHg.⁴⁶ The similarity of retinal and vitreal PO₂ may break down under some conditions, however, including diabetes mellitus.⁴⁷

The values of inner retinal PO₂ cover a relatively wide range(Figure 3). The average PO₂ is a few millimeters of mercury lower in light adaptation.^22,45 The O₂ extraction from the retinal circulation is high relative to that in many tissues, about 8 mL O₂ per 100 mL blood, ¹⁰ resulting in a low venous PO₂ that contributes to making the average tissue PO₂ in the retina lower than in many organs.⁴⁸

Oxygen consumption can be derived from material balances in vivo from blood arteriovenous differences^49,50 or in vitro from reductions in PO₂ in a closed chamber.^3,51,52 Oxygen consumption can also be obtained by fitting mathematical models of O₂ diffusion and consumption to PO₂ profiles.^22,26,53 The use of material balances does not provide information on the spatial distribution of consumption, but was the first method to show that QO₂ was higher during dark than light adaptation in the whole retina.^52,54 Further aspects of retinal energy metabolism, particularly the unusual amount of glycolytic metabolism, have been reviewed recently.^3,55

Absolute values of photoreceptor QO₂ were computed by fitting a mathematical model of O₂ diffusion in the outer retina to the O₂ profiles.⁵³ In the most satisfactory model, all the O₂ use is confined to the inner segment layer. It is important to realize that even though QO₂ is negligible in the outer segment and outer nuclear layers, linear gradients of O₂ still remain throughout these regions (Figure 2, Figure 4, and Figure 5). The evidence of a nonzero QO₂ is not the presence of a gradient but curvature in the profile, specifically a quadratic dependence of PO₂ on distance.⁵³

Table 1 shows that QO₂ of the photoreceptors during steady illumination is 36% to 68% of the value in the dark. This agrees with the effect of light found by other methods in the rat, rabbit, and pig.^6,56,58 Above rod saturation, which is below-normal room illumination, no further changes in QO₂ are observed in the rod-dominated area centralis of the cat retina.²² Modeling also shows that the choroid is essential to photoreceptor metabolism, contributing approximately 90% of the O₂ consumed by the photoreceptors in darkness, and all of the O₂ consumed during light adaptation.⁴⁵ Finally, the choroid contributes approximately 50% of the O₂ consumed by the whole retina in dark adaptation, ^26,56 which is consistent with earlier estimates based on choroidal arteriovenous O₂ differences and blood flow.³⁹

Table 1 gives values for QO₂ as though the consumption were uniform across the whole outer retina. We have preferred to report the values this way, because it reduces the scatter in the measurements⁵³ and makes values from modeling more comparable to those obtained by other methods. These values are somewhat misleading, however, since the QO₂ is localized to the inner segments. The rod inner segments have an extremely high QO₂ of 15 to 20 mL O₂ (standard temperature and pressure) per 100 g of tissue per minute in the dark, consistent with the high density of mitochondria in this region.

Little is known about cone QO₂. Light does decrease the QO₂ of monkey fovea.³⁵ Cones in the monkey have a volume of mitochondria that is about 10 times that of rods, ⁵⁹ but we have argued that QO₂ of an individual cone could not be higher than that of a rod by a factor of 10, because the smaller number of mitochondria in rods already uses all the O₂ available from the circulation.

Oxygen is used in photoreceptors for many processes, but the ones that appear to require the most energy are maintenance of the dark current, ie, pumping out the large load of sodium ion that enters through the light-dependent channels^54,60,61 and generating guanosine triphosphate, from which cyclic guanosine monophosphate is produced.^3,61 The dark current decreases dramatically during illumination, whereas the guanosine triphosphate turnover increases. These appear to be the only processes whose metabolic demands change significantly with illumination.⁶¹ The change in sodium ion pumping is larger, however, and, as a result, illumination causes a net decrease in photoreceptor QO₂ in all species investigated to date.

In the inner retina, the retinal circulation forms a 3-dimensional mesh, which makes mathematical modeling difficult. The retinal circulation must be occluded to allow the application of a diffusion model to the inner retina.^26,62 Despite this limitation, various lines of evidence indicate that in the inner retina, no difference in QO₂ exists between darkness and steady illumination^26,56,63 (Table 1). The metabolism of the inner retina is higher when flickered light (4 Hz) is applied as opposed to steady light, ⁶³ presumably because flickered light leads to greater neural activation. Unfortunately, this increase is difficult to quantify.

Figure 3 shows intraretinal O₂ profiles from the dark-adapted cat retina during normoxia and hypoxemia, induced by adding nitrogen to the inspired gas. In the inner retina, tissue PO₂ is effectively regulated across a wide range of Pa O₂. The decrease in preretinal PO₂ is only 0.14 to 0.18 mm Hg/mm Hg PaO₂ when the PaO₂ is above 35 mmHg^28,64 in the cat, and 0.2 mm Hg/mm Hg PaO₂ in the miniature pig.⁶⁵ When the PaO₂ falls below 35 mmHg, PO₂ in the inner retina decreases more steeply, approximately 0.62 mm Hg/mm Hg PaO₂.⁶⁴ The good regulation at PaO₂ of greater than 35 mmHg results from vasodilation of the retinal circulation, ^66,67 which in cats can increase blood flow by more than a factor of 3 during hypoxemia.⁶⁸

In the outer retina, hypoxemia leads to a steep decrease in choroidal PO₂ (Figure 4) of approximately 0.64 mm Hg/mm Hg PaO₂.⁴⁵ This occurs because the choroidal blood flow does not increase in hypoxemia⁶⁹ (Melissa Pulfer, Jameel Ahmed, PhD, and R.A.L., unpublished data, June 2000). There would be little advantage of increasing choroidal blood flow, even if this were possible. The usual advantage of increasing blood flow to a tissue during hypoxemia is allowing the arteriovenous O₂ difference to be reduced, which minimizes or prevents a decrease in venous PO₂ during hypoxemia. In the choroid, the arteriovenous O₂ difference is already very small, only about 1 volume percent, ²⁸ due to the high rate of blood flow. In relatively mild hypoxemia, arterial PO₂ drops below the normoxic choroidal venous PO₂. Reducing the arteriovenous difference could not prevent this. Because photoreceptor metabolism is O₂ limited under normoxic conditions in darkness, even mild hypoxemia (PaO₂ of 60 mmHg) reduces the flux of O₂ to the inner segments and, therefore, reduces photoreceptor QO₂.⁴⁵ When hypoxemia occurs during light adaptation, choroidal PO₂ also falls, but in this case, metabolism is not O₂ limited, so photoreceptor QO₂ is not affected.

When the intraocular pressure (IOP) is elevated, the effects on retinal PO₂ are similar to those observed during hypoxemia.^70,71 Decreased perfusion pressure does not affect PO₂ in the inner retina owing to effective autoregulation in the retinal circulation.^39,72,73 In contrast, when the IOP is elevated, the choroidal blood flow decreases, ^72,74-76 which leads to a reduction of choroidal PO₂ of approximately 0.5 mm Hg/mm Hg of perfusion pressure.⁷⁰ This reduction leads to a reduction in photoreceptor QO₂.⁷⁰ An extension of this finding is that anything that decreases choroidal blood flow would be expected to have a negative impact on retinal oxygenation and on photoreceptors. There is evidence that human and experimental glaucomas affect photoreceptors, ⁷⁷ possibly because they reduce photoreceptor O₂ supply, and there is direct evidence of retinal damage by reduced choroidal blood flow in birds.⁷⁸

Figure 5 compares a normoxic profile to one obtained during hyperoxia (inspiration of 100% O₂).Vitreal PO₂²⁸ and inner retinal PO₂^{29,30,32,57,79} are well regulated during hyperoxia in the cat, pig, and rat and presumably in humans. This regulation is due to constriction of retinal vessels^10,66,80 and reduction of retinal blood flow.^10,66,81 In the miniature pig, PO₂ in the inner retina has been reported to be perfectly regulated during hyperoxia in intervascular zones, ^29,32 but it increases by an average of approximately 40 mmHg in juxta-arteriolar zones, the same value observed in cats.^30,57 Adding 5% carbon dioxide to the inspired gas leads to larger increases in inner retinal PO₂ during hyperoxia.¹⁶ The most striking change during hyperoxia is the increase in choroidal PO₂ to about 250 mmHg in cats³⁰ and 220 mmHg in miniature pigs.³² This increase is a consequence of the lack of metabolic regulation of choroidal blood flow.⁸²

It is sometimes thought that hyperoxic therapy has a limited potential for treating retinal disease. This is based on the idea that the blood is nearly saturated with O₂ during normoxia, so that the small additional O₂ content during hyperoxia should be of little benefit. The fundamental misconception is that the O₂ supply to the tissue relies on O₂ saturation, whereas, in reality, O₂ moves into tissue by simple diffusion, which is driven by gradients of PO₂, not saturation. The PO₂ gradient from the choroid is much steeper during hyperoxia, and a greater portion of the retina can be supplied by the choroid during hyperoxia than during normoxia, ^30,32 as shown in Figure 5.

Despite an abundance of O₂, higher photoreceptor QO₂ was not found in cat retina during hyperoxia.^30,57 Pournaras and coworkers^32,83 argued that QO₂ increased during hyperoxia in healthy retinas of the miniature pig, but not in retinas with a branch vein occlusion. Their conclusions were not based on calculations of QO₂, however, but only on the steepness of gradients, which cannot be used to infer values for QO₂. The gradient through the outer segments is steeper during hyperoxia, so the flux across them is larger, but they never use any O₂.

For a few hours, at least, hyperoxia does not impair retinal electrical function in animals, ⁵⁷ but the issue of O₂ toxicity to the retina has not been fully addressed. It is clear that O₂ can be damaging to photoreceptors if combined with high levels of illumination.^84,85

Diabetic retinopathy is a disease predominantly of the retinal vasculature, leading first to capillary occlusion and then to vascular proliferation. For many years, tissue hypoxia has been suggested to be involved in the progression of diabetic retinopathy⁸⁶ and retinal neovascularization in general.⁸⁷ By the time that capillary nonperfusion is clinically observable, tissue hypoxia has likely occurred, but the exact time when hypoxia begins is as yet unknown. Retinal tissue hypoxia was not found in the early stage of diabetic retinopathy in animal models.^46,88,89 However, intraretinal O₂ measurements on cats with long-term (6-7 years) diabetes mellitus showed an abnormally low average inner retinal PO₂ of 7.7 mmHg(compared with 16.4 mmHg in normal cats).⁴⁷ This report, the only one to provide intraretinal measurements, stressed that tissue hypoxia was present in the diabetic retina before capillary dropout was evident and when only a few microaneurysms were present. Further evidence of hypoxia relatively early in the disease was provided by the observation that hyperoxia reversed early contrast sensitivity deficits⁹⁰ and oscillatory potential reductions.⁹¹ In addition, Arden et al⁹² provided evidence of altered dark adaptation in diabetes mellitus, which was also suggested to be an effect of hypoxia.

Hypoxia may occur during diabetic retinopathy due to capillary occlusion. Leukocytes have been reported to be present in greater number and are less deformable in diabetic cats.⁹³ Increased levels of endothelial cell/leukocyte adhesion molecules (intercellular adhesion molecule 1 and P-selectin) have also been observed in humans with diabetes mellitus(hereafter referred to as diabetic humans).⁹⁴ These findings may explain the increase of leukocyte activation and adhesion in diabetic humans⁹⁵ and rats, ⁹⁶ which are known to increase vascular resistance^97,98 as capillaries become plugged. In addition, a hyperglycemic (hyperosmotic) environment increases the adhesion of leukocytes to retinal endothelial cells.⁹⁹

The findings of tissue hypoxia and occlusion of retinal capillaries by leukocytes are consistent with observations that retinal blood flow is reduced before and in the early stages of diabetic retinopathy.^100-103 A subsequent increase in retinal blood flow occurs, ^102,103 possibly because lower-resistance shunt pathways develop and carry some of the flow¹⁰² and/or because vascular endothelial growth factor(VEGF) dilates vessels.¹⁰³ The flow is reduced again in proliferative retinopathy.^104,105 Complete consensus on this sequence of events has not been achieved, but they seem to explain most of the observations.

One of the puzzles about diabetic retinopathy has always been why it affects the microvasculature of the retina more than that of the brain.¹⁰⁶ Arden et al⁹² suggested that dark adaptation aggravates hypoxia by depriving the inner retina of the small amount of O₂ that diffuses from the choroid during light adaptation. They suggested further that avoiding a long period of dark adaptation(eg, during sleeping at night) could be an alternative therapy for diabetic retinopathy.

Hypoxia is thought to induce the synthesis of VEGF, ^107,108 which is present in higher levels during proliferative diabetic retinopathy.^109,110 Vascular endothelial growth factor has also been found in eyes with no retinopathy or only early signs of retinopathy.^111-113 This initially suggested that factors besides hypoxia may induce the production of VEGF, ¹¹² but we now know that the retina is hypoxic early in the disease.⁴⁷ Although VEGF is important, other angiogenic^114-118 and antiangiogenic^119,120 factors may also be involved. A detailed discussion on factors involved in intraocular neovascularization is available elsewhere.¹²¹ A competing theory of vasoproliferation implicated the mechanical aspect of vasodilation resulting from hypoxia or hyperglycemia.¹²² This theory has received little attention recently, but it seems prudent to recognize that mechanical factors could play a role in angiogenesis.

The QO₂ of the retina in diabetes mellitus has been a subject of controversy. In the intact cat retina, photoreceptor QO₂ was lower on average in subjects with diabetes than in healthy subjects, ⁴⁷ but QO₂ has a strong positive correlation with choroidal PO₂, and the choroidal PO₂ values tended to be lower in diabetic cats. Whether lower choroidal PO₂ is a general feature in diabetes could not be determined with the number of animals available in that study. In vitro studies in rabbits suggested that the QO₂ decreased in diabetes.^123,124 In contrast, one study reported that QO₂ in diabetic rats¹²⁵ was greater. Little work has been done on the QO₂ of the inner retina. Tiedeman et al¹² found that retinal venous O₂ saturation was lower than normal in diabetic humans, and concluded on that basis that QO₂ of the inner retina must be greater in subjects with diabetes. If blood flow increased or was unchanged, this would be a reasonable conclusion, but blood flow was not measured in that study. Increased QO₂ is difficult to reconcile with capillary dropout and inner retinal damage.

On the theory that one of the major problems in diabetic retinopathy is lack of O₂ in the inner retina, therapies have been designed to improve the O₂ supply. Panretinal photocoagulation is effective in treating proliferative retinopathy.^126-130 Studies on animal models^46,131-134 and in humans²⁴ showed that PO₂ in the preretinal vitreous is higher over photocoagulated regions than over normal regions. These measurements provided little detail about O₂ gradients in the retina, and were obtained mainly during hyperoxia, which is not the condition in which diabetic retinopathy develops or progresses. Nevertheless, they provide some evidence that photocoagulation reduces the QO₂ of the photoreceptors, allowing more O₂ to diffuse from the choroid to the inner retina.^135,136

At present, no treatment stops retinopathy apart from panretinal photocoagulation, but it seems reasonable that if early retinal circulatory changes preceding VEGF up-regulation could be detected clinically, then interfering with leukocyte adhesion or using other pharmacological techniques to increase retinal blood flow might be effective in preventing retinopathy.

Retinopathy of prematurity (ROP), previously known as retrolental fibroplasia, ¹³⁷ develops in 3 distinct stages.¹³⁸ First, the retina of premature infants is exposed to excess O₂ as a result of breathing O₂-enriched gas, which causes constriction and irreversible closure of retinal vessels and prevents new vessels from developing normally.^139,140 As a result, hypoxia is thought to develop in the inner retina when the infant is returned to breathing air, and the extra supply of O₂ from the choroid is lost. Finally, this hypoxic condition induces vascularization to counteract the hypoxia, but for some reason this vascularization is abnormal.^140-142

The current therapy for prevention of the disease is generally to titrate the infant's blood to an adequate PaO₂ but to prevent systemic hyperoxia. Moderate supplemental O₂ with a target arterial saturation of 99% and a PaO₂ of less than 100 mmHg was found to decrease prethreshold ROP.^143-145 Vitamin E supplementation may also be useful in decreasing the severity of ROP, ^146-148 possibly by minimizing auto-oxidative reactions that arrest the growth of normal blood vessels.¹⁴⁸ Other therapies, such as gradual reduction in O₂ level (oxygen weaning) appear less promising.^149-151

Mechanistically, the disease is not as straightforward as the above discussion would suggest. As in diabetic retinopathy, it has been impossible to determine exactly what the oxygenation status of the human retina is at any time during the disease progression, and difficult even in the animal models. Ernest and Goldstick¹⁵² used O₂ microelectrodes in the vitreous humor of kittens after O₂ supplementation(80%-90% O₂) and found that preretinal PO₂ over the avascular retina was close to 0 mmHg, but normal over the vascular retina. These remain the only direct measurements of PO₂ in an ROP model. More recently, attention has shifted to a rodent model of ROP called ischemia-induced or O₂-induced retinopathy.^153,154 Retinopathy is produced most effectively by a regimen that cycles the inspired gas between a relatively hyperoxic one and a relatively hypoxic or normoxic one for at least several days during retinal vascular maturation.¹⁵⁴ Surprisingly, well-controlled hyperoxia followed by normoxia produces ROP-like symptoms with less certainty than does O₂ cycling. In O₂-induced retinopathy, some evidence from novel magnetic resonance imaging measurements suggests that hypoxia plays a role early in the disease.¹⁵⁵ Berkowitz and Penn¹⁵⁵ based this conclusion on a blunted response of retinal oxygenation to breathing of carbogen (a mixture of 95% O₂ and 5% carbon dioxide) in animals in which histological signs of the disease developed, rather than a direct demonstration of tissue hypoxia.

As in diabetic retinopathy, it is thought that hypoxia does not directly affect vessel growth, but instead influences the production of angiogenic factors in the retina, ¹³⁸ with the probable involvement of hypoxia-inducible factor 1 as an intermediate.¹⁵⁶

Work on complete ocular ischemia, produced in general with elevation of IOP, and work on the effects of anoxia on the retina in vitro have a long history. Although they are relevant to ocular ischemic syndrome, in which the retinal and choroidal circulations are compromised by carotid artery obstruction, ¹⁵⁷ they will not be reviewed herein. Of more interest from a clinical standpoint is the role of O₂ in central and branch venous and arterial occlusions of the retinal circulation. Experimental venous occlusion in pigs leads to decreased inner retinal PO_2, which can be followed by neovascularization.¹⁵⁸ Work on cats shows that retinal arterial occlusion makes the entire inner retina anoxic, ^57,61 and this causes a rapid failure of the ERG b wave in cats and monkeys.^57,159,160 Even when the animal breathes air, the retina is reasonably tolerant to occlusion, surviving episodes of 100 to 120 minutes with full recovery of the ERG^159,160 after the occlusion is reversed. Several studies have shown that retinal oxygenation can be partially or completely restored during the arterial occlusion by making animals hyperoxic.^{57,61,161,162} It is clear from O₂ profiles that this effect relies on an increased O₂ supply from the choroid during hyperoxia. Corroborating the importance of O₂ as the limiting factor during occlusion are a few studies that have used a model of total ischemia. Anderson and Saltzman² have shown that if human subjects breathed O₂ before IOP elevation, their vision was sustained longer than if they breathed air. Blair and coworkers¹⁶³ have shown that perfusing the vitreous with an oxygenated solution after total occlusion can maintain the structural integrity of the retina and the ERG. Despite this positive experimental data, the clinical experience with hyperoxia has been mixed, with only a few studies recommending hyperoxia as a treatment for vascular occlusion.^164,165 Unfortunately, most clinical attempts to provide hyperoxia during occlusion have not adequately considered that it must be provided for large blocks of time and as soon as possible because O₂ is not stored and is used so rapidly. In addition, there have been two fears of using hyperoxic therapy. First, hyperoxia constricts retinal vessels, and this might impede the clearance of emboli. This may be an unnecessary concern, because the inner retina is acidic during occlusion(Gulnur Birol, PhD, N.D.W.-W., Ewa Budzynski, MS, and R.A.L., unpublished data, October 2002), which would tend to counteract hyperoxic vasoconstriction. In addition, carbon dioxide can be added to the inspired gas to induce vasodilation.^164,165 Second, extended O₂ breathing can be toxic, but this should not prevent hyperoxic therapy, because it is possible to breathe 60% to 70% O₂ for hours without toxic effects.¹⁶⁶

Oxygen undoubtedly plays a role in other diseases. Several studies report that cystoid macular edema can be ameliorated by hyperbaric O₂ therapy.^167-170 No clear theoretical reason for this exists. Depending on the mechanism of this effect, hyperoxia at normal atmospheric pressure may do just as well. This has not been tested, but should be, because it would allow application of the therapy to more patients.

Oxygen is likely to play a role in the dramatic vasoattenuation of the retinal circulation that is associated with photoreceptor degenerations in humans and animals.¹⁷¹ This role can be understood on the basis of the normal gradients. When photoreceptors are lost, O₂ derived from the choroid that was once used by photoreceptors now diffuses all the way to the inner retina.^172,173 In addition to the reduced size of the major vessels, capillaries in the inner retina are permanently lost¹⁷⁴ in animals with photoreceptor degeneration. This capillary loss can be prevented by maintaining the animals under hypoxic conditions, which probably reduce the flux of O₂ from the choroid to the inner retina. Evidence also exists that O₂ can modulate the survival of photoreceptors early in the Royal College of Surgeons rat model of retinal degeneration. Hyperoxia is protective at early stages, whereas hypoxia induces death of photoreceptors.¹⁷⁵ Up-regulation of basic fibroblast growth factor by hypoxia is not sufficient to prevent photoreceptor death, ¹⁷⁶ and later stages of the photoreceptor loss may not be caused by hypoxia.

Although a focus on mechanical causes of ganglion cell loss in glaucoma continues, many lines of evidence indicate that reduced ocular blood flow, and by implication reduced O₂, is a contributing factor in the retinal damage in glaucoma.^177,178 This is a complex area in which most of the evidence is indirect, but support for a vascular hypothesis continues to accumulate. First, hyperbaric O₂, which does not change IOP, can improve visual fields in glaucomatous subjects.¹⁷⁹ Second, photocoagulation, which destroys photoreceptors and presumably increases the inner retinal O₂ supply from the choroid, promotes survival of ganglion cells in glaucomatous monkeys.¹⁸⁰ Third, cones, which should not be affected directly by mechanical stresses at the optic nerve, are swollen in eyes of humans with glaucoma and in primates with experimental glaucoma.¹⁸¹ Finally, systemic vascular variables, which have been known to be risk factors for glaucoma, help predict the degree of visual dysfunction in glaucoma.¹⁸²

Oxygen may also play a role in vision loss in retinal detachment. A recent argument has been made on a theoretical basis that hyperoxia may be useful in preventing photoreceptor damage.⁴¹ Detachment separates the inner segments from their O₂ supply, and although no consuming tissue is found under the retina, the increased distance reduces the flux of O₂ from the choroid to the photoreceptor inner segments. Hyperoxia should restore this flux, at least for detachments of moderate height, because increased choroidal PO₂ can compensate for the increased distance. For large detachments, the photoreceptors may benefit more from increased amounts of O₂ in the retinal circulation than in the choroidal circulation. The protective effect of hyperoxia on detached photoreceptors has been shown experimentally in cats, in which hyperoxia was able to save photoreceptors and prevent the activation of retinal glia normally caused by detachment.^183,184

Most of the important features of normal retinal oxygenation are now known, and the status of retinal oxygenation in some prevalent diseases has been partially elucidated. Tissue hypoxia is believed to be important in regulating vascular growth factors, particularly in the vasoproliferation observed in diabetes mellitus and ROP. Hypoxia is also involved in the loss of photoreceptors during a retinal detachment and the loss of inner retinal neurons in retinal arterial occlusions. Evidence is emerging that hypoxia may damage photoreceptors and possibly other cells in glaucoma, and that O₂ modulates the survival of photoreceptors in photoreceptor degenerations. One of the most interesting aspects of retinal oxygenation is the important metabolic role of the choroidal circulation. Lack of regulation of the choroid by O₂ may be detrimental to the photoreceptors under conditions when choroidal PO₂ or choroidal blood flow decrease (eg, glaucoma), but this same lack of regulation provides an opportunity to use the choroid therapeutically in treating cystoid macular edema, retinal arterial occlusions, and retinal detachment. Oxygen therapy only has a chance of working if it is sustained, however, and this has generally not been the mode in which it has been used.

The difficulty of measuring O₂ intraretinally in humans or in animal models of human diseases has prevented a more complete understanding of the role of O₂ in retinal diseases. Measuring retinal O₂ would be far better than measuring blood flow, because tissue PO₂ is the variable that is directly relevant to metabolism and presumably to visual function. Noninvasive measurements have been difficult, because of the spatial heterogeneity of PO₂ in the retina and the difficulty of accessing the choroid, but these techniques are important and need to be developed. Development of noninvasive optical techniques would allow one to study the timing of hypoxia with respect to biochemical and histological changes, and would serve as a way to follow disease progression and evaluate treatments.

Corresponding author and reprints: Robert A. Linsenmeier, PhD, Department of Biomedical Engineering, Northwestern University, 2145 Sheridan Rd, Evanston, IL 60208-3107 (e-mail: r-linsenmeier@northwestern.edu).

Submitted for publication January 4, 2002; final revision received November 15, 2002; accepted December 6, 2002.

This study was supported by grant EY05034 from the National Institutes of Health, Bethesda, Md.

We thank Lissa Padnick-Silver, PhD, for useful feedback.