Ординатура / Офтальмология / Английские материалы / Retinal and Choroidal Angiogenesis_Penn_2008

existing circulation through the inner limiting membrane and into the vitreous and are subsequently associated with vision loss and blinding complications in retinopathies such as diabetic retinopathy and retinopathy of prematurity (ROP). This ocular pathological angiogenic process is termed neovascularization (NV).

The development of all three forms of retinal vessels is commonly thought to occur when oxygen supply is inadequate to meet demand during resting conditions, or hypoxia. The retina is one of the most metabolically active tissues in the body, and it has a very high oxygen demand.1 Because oxygen is not stored within retinal tissue, a continuous supply of oxygen is necessary to maintain adequate retinal nutrition. Consequently, oxygen supply and demand must be precisely balanced through active regulation of nutrient delivery and waste removal to ensure the health of the retina. Retinal hypoxia can occur, for example, during a retinal ischemic event (e.g., branch retinal artery occlusion) in which oxygen supply stops but oxygen consumption is not downregulated.

Historically, this hypoxia hypothesis (Figure 1) evolved from the initial work of Michaelson in 1948. He studied excised, india ink-injected preand postnatal retinas and noted that there were large capillary-free zones around arteries and that capillary growth tended to occur on the side of the vein farthest from the artery.2 Presumably, these avascular regions were receiving adequate amounts of oxygen from arteries relative to demand. Michaelson hypothesized that in the development of embryonic retinal vessels, and possibly for NV, oxygen concentration gradients from well to poorly oxygenated retina (e.g., from artery to vein) regulated a factor “X,” which in turn influenced new vessel development. Furthermore, subsequent work by others showed that as the inspired oxygen fraction increased or decreased, the size of the capillary-free zones around arteries widened or narrowed, respectively.3 In support of Michaelson’s hypothesis, Chan-Ling et al. examined normal retinal vessel development in kittens by measuring the extent of vasculogenic cell division.4 Cell division was found to be inversely proportional to the level of oxygen in the inspired gas mixture, and they speculated that a “physiological” level of hypoxia related to increased retinal neuronal demand stimulates vasculogenesis.4

Figure 8-1. Schematic of hypoxia hypothesis: a) during normoxia, no growth of new blood vessels is noted, b) during hypoxia, retina sends out biochemical signals which induce NV from existing circulation and, in turn, brings oxygen to relieve hypoxic region (and removes waste products).

To date, the hypoxia hypothesis has been applied to rationalize both normal retinal vessel growth and the development of retinal NV in diseases such as diabetic retinopathy, age-related macular degeneration, retinopathy of prematurity, and sickle cell retinopathy. However, it has remained unclear how hypoxia alone can be a necessary and sufficient condition (i.e., a cause) for the phenotype change to NV as well as normal vasculogenesis and angiogenesis. This chapter will critically examine the evidence for and against the hypoxia hypothesis. As will be seen, a causative link between hypoxia and NV is not strongly supported.

2.HYPOXIA AND RETINAL NV: PROS AND CONS

In general, only indirect evidence (such as treatment response in patients and biochemical and oxygen measurements in animal models) has been used to support the hypothesis that hypoxia causes retinal NV. Our working definition of hypoxia is an inadequate supply of oxygen relative to demand.

In this discussion, oxygen measurements are considered indirect because retinal oxygen demand (i.e., consumption) is not measured but is needed, according to our definition, to formally make an assessment of hypoxia. Oxygen consumption measurements in vivo are difficult, and it is often assumed that consumption changes remain small between control and experimental conditions.

3.TREATMENT RESPONSE

If we assume for the moment that retinal hypoxia does cause NV, then one obvious treatment for minimizing NV would be to alleviate poor oxygenation by administering supplemental oxygen. In fact, experimental studies of proliferative retinopathy have demonstrated that constantly applied supplemental oxygen significantly reduces the risk of developing experimental retinal NV.5-7 These results helped motivate the National Eye Institue-sponsored multicenter clinical trial (STOP-ROP) to test the efficacy, safety, and costs of providing oxygenation in moderately severe prethreshold ROP.8 However, the STOP-ROP trial did not demonstrate that supplemental oxygen produced a significant reduction in the number of infants requiring peripheral ablative surgery for retinal NV compared with conventional oxygen exposure.9 In other words, the STOP-ROP trial did not demonstrate a beneficial effect of supplemental oxygen on NV.

Although there are likely several reasons why the STOP-ROP results were not as expected, we wondered if one possibility was that administering oxygen 100% of the time (i.e., constantly) was impractical due to the daily care needs of at-risk neonates and that instead infants experienced a variable supplemental oxygen exposure. This hypothesis was tested in a simple model of variable supplemental oxygen in which oxygen was administered only 99% of the time.10 It was expected that supplemental oxygen administered either 100% or 99% of the time would reduce retinal hypoxia by similar degrees and thereby lessen retinal NV to comparable extents. Instead, we found that variable supplemental oxygen treatment was significantly less effective at reducing retinal NV than constantly applied supplemental oxygen.10 This outcome was somewhat surprising and raised the possibility that the beneficial effects of constantly applied supplemental oxygen on retinal NV are not entirely due to relieving retinal hypoxia.

An alternative explanation may be that supplemental oxygen induces some degree of vasoconstriction, perhaps through the expression of endothelin-1, and this in turn alters retinal perfusion patterns.11 This hypothesis has not yet been tested because, to the best of our knowledge, it has been difficult to accurately quantitate retinal perfusion clinically or in

animal models of retinal NV during room air and supplemental oxygen breathing. Current noninvasive methods for measuring retinal perfusion are either not quantitative (e.g., fluorescein angiography), have limited spatial resolution and sensitivity (e.g., laser Doppler velocimetry), or are limited by media opacities such as cataracts or the presence of hyaloidal circulation (e.g., video fluorescein angiography). Nonetheless, the results from the supplemental oxygen studies do not appear to unequivocally demonstrate a link between hypoxia and NV.

Laser treatment is a useful clinical procedure that has been found to minimize harmful visual consequences of retinal NV in, for example, patients with diabetic retinopathy, sickle cell retinopathy, and ROP. In principle, laser treatment, which destroys small retinal regions, can reduce oxygen consumption by the retinal pigment epithelium (RPE)-photoreceptor complex, thereby increasing oxygen availability from the choroidal circulation to the inner retina.12 However, data supporting an association between laser procedures and increased oxygen availability have been somewhat weak. Experimentally, laser treatment has been found to elevate retinal oxygen levels during room air breathing in nearly avascular retina (rabbit) but not in fully vascularized retina (cat).13,14 Zuckerman, Cheasty, and Wang measured increased inner retinal pO2 over laser burn in rats, but it was unclear whether their data were obtained during room air or 100% oxygen breathing.15 Clinically, some patients with diabetes, despite extensive laser treatment that would be expected to improve hypoxia, have NV that continues to develop and cause complications. Why laser treatment is beneficial in reducing the impact of retinal NV is unclear at present but may involve changes in retinal perfusion.16 For example, it has been reported that laser treatment produces a prolonged decrease in retinal perfusion,17,18 and this altered perfusion pattern might affect the NV outcome independently from whether hypoxia is present or not.

In summary, adverse consequences of retinal NV can be reduced to a variable and somewhat unpredictable extent using laser treatment, but whether or not this benefit occurs by increasing oxygen availability and reducing a presumed hypoxia remains speculative. In addition, the reasons for the inconsistent improvement in outcome following these clinical approaches are unclear but may be related to non-hypoxic factors, such as changes in perfusion patterns17,18 or vasoreactivity (see below) and/or individual inflammatory response to the procedures.19 In any event, treatment responses are likely to result from a complex set of mechanisms, and their interpretation in terms of cause and effect between hypoxia and NV remains tentative.

4.BIOCHEMICAL EVIDENCE

The most likely candidate yet identified for Michaelson’s factor “X” is vascular endothelial growth factor (VEGF).20 VEGF is a potent hypoxiainducible mitogen found upregulated in cell cultures as well as in proliferative retinopathy. However, in enucleated eyes from patients with diabetes, evidence for retinal VEGF immunoreactivity has also been found before the appearance of gross retinal nonperfusion. While normal histology may suggest normal retinal oxygenation (i.e., normoxia), some caution is needed because retinal hypoxia was found in long-term diabetic cats that also presented with relatively mild histopathology.21 Nonetheless, the data from enucleated eyes at least supports the possibility that hypoxia may not be the sole stimulus for VEGF expression.22,23 Indeed, a variety of factors unrelated to hypoxia, but apparently related to cellular malnutrition, can also upregulate VEGF expression including, for example, increased insulin-like growth factor-1 (IGF-1) and oxidative stress.24,25 These observations raise some reservations about interpreting elevated VEGF levels solely in terms of hypoxia.

The appearance of NV likely depends on more than an increased level of a single factor “X.” A minimum requirement for NV development may be that more angiogenic stimulators (e.g., VEGF) are present than angiogenic inhibitors (e.g., pigment epithelium-derived factor (PEDF)).26 Interestingly, hypoxia appears to suppress PEDF expression, and it has been suggested that PEDF downregulation is linked with the development of retinal NV, at least in animal models.27,28 However, this suggestion has not been supported by the available evidence since measures of PEDF levels in vitreous of patients with proliferative retinopathy have been reported as both higher and lower than normal.29

Changes in systemic biochemistry that are apparently unrelated to retinal hypoxia have also been strongly linked to the development of proliferative retinopathy. It has been suggested that metabolic acidosis is an independent risk factor associated with clinical ROP.30 Experimental support for this supposition has been established by the work of Holmes et al., who report that lowering systemic pH using either carbon dioxide, ammonium chloride, or acetazolamide, is a key factor in inducing retinal NV in newborn rats independently of hyperoxemia or hypoxemia.30-32

Other studies have recognized a link between aspects of thyroid function (as assessed by serum levels of thyroxine (T4), thyroid-stimulating hormone (TSH), and IGF-1) and retinal hemodynamics, endothelial cell barrier function, and NV formation.33-36 Neonates born prematurely (e.g., at week 25 or earlier) can present with very low birth weight, lower than normal T4 concentration, and an incompletely developed retinal circulation and large

avascular (ischemic) sections of peripheral retina. In addition, the development of ROP in very low birth weight infants has been associated with prolonged periods of low plasma IGF-1 levels.37 Lack of IGF-1 in knockout mice prevented normal retinal vascular growth.37 In our animal studies, methylimidazole-induced hypothyroidism in both control and a model of low retinal NV incidence newborn rats confirmed that thyroid function is linked with normal retinal vascular density and that hypothyroidism can play a permissive role in the development of retinal NV, if some risk of NV already exists.38

In summary, changes in retinal biochemistry alone have not unambiguously proven the hypoxia hypothesis.

5.OXYGEN MEASUREMENTS

5.1Methods

Several methods have been proposed to measure retinal oxygen tension including oxygen microelectrodes and spectroscopy (e.g., optical and 19F NMR). The most frequently used method of determining retinal oxygen tension (pO2) is through the use of oxygen-sensitive electrodes. This technique can be used to measure pO2 gradients within local regions of the vitreous and/or retina with high spatial and temporal resolution. Intravitreal and intraretinal oxygen electrode studies in vascularized retina, such as the cat or rat, have revealed that the vitreous (which is avascular) does not consume oxygen and that during normoxia, oxygen emanating from the choroidal circulation is effectively prevented from reaching the inner retina and vitreous by the high oxygen consumption rate of the photoreceptors.1,39 For this reason, and because there is a small diffusion distance between retinal vessels and vitreous/retina, oxygen electrode studies have shown that measurement of vitreous oxygen and its changes near the surface of the retina (i.e., preretinal) accurately mirror oxygenation of the most anterior portion of inner retina.1,39 Unfortunately, the invasive nature of this technique is a major limitation. To date, clinical measures of retinal O2 in the preretinal vitreous using an oxygen electrode have only been made on patients undergoing intraocular surgery and have not been useful in addressing questions of hypoxia and retinal NV.40,41

Less frequently used approaches to measuring retinal oxygen tension involve spectroscopic techniques. Several groups have developed methods for estimating retinal arterial and venous oxygen saturation that involve detecting the difference in light absorption between oxygenated and

deoxygenated hemoglobin using multiple wavelength reflectance oximetry.42,43 This method is expected to be most useful for studying retinopathy associated with retinal hypoxia. However, a major technical limitation that remains to be overcome is the high degree of variability induced by backscattered light, which can confound data interpretation. Other groups have measured intravascular or perivascular pO2 in animal models using methods based on the quenching of phosphorescent or fluorescent dyes by oxygen.15,44,45 This method requires careful attention to light intensities, because at some light levels, the dyes used can produce toxic oxygen free radicals. Furthermore, additional work is needed to unambiguously separate the optical signals that are derived from the retinal and choroidal circulation. 19F magnetic resonance spectroscopy of a perfluorocarbon droplet also has been used to measure preretinal pO2, both in animal models and in a vitrectomized human eye.46-49 This approach is minimally invasive because the droplet is delivered via a 30 g needle, but the technique reflects the retinal oxygen tension with good spatial resolution (only where the droplet is). A major advantage of this approach is that it can measure preretinal oxygen tension under conditions that the previously mentioned methods cannot, such as in newborn rat eyes, mouse eyes, or eyes without a clear optical medium.49,50

5.2Results

To the best of our knowledge, relatively few studies have attempted to test the hypoxia hypothesis by measuring retinal oxygen levels in models of proliferative retinopathy.46,49,51-53

Pournaras found, in an experimental retinal branch vein occlusion, achieved by using argon laser photocoagulation in miniature pigs, that the preretinal vitreous over all ischemic foci had subnormal pO2 (measured with an oxygen electrode) before the appearance of NV, but only approximately 45% of these ischemic retinas showed development of NV.51 He speculated that a critical level of hypoxia was needed for NV formation. An alternative possibility is that retinal hypoxia was a necessary but not sufficient factor leading to NV. In any case, it is clear that hypoxia alone was not correlated with NV growth in the pig occlusion model.

Ernest and Goldstick measured preretinal vitreous oxygen tensions using microelectrodes in a kitten oxygen-induced retinopathy model.52 In this model, the retinal blood vessels of newborn kittens were largely obliterated by exposure to an atmosphere of 80% to 90% oxygen, producing what could be considered a vascular wound. Avascular retina was found to indeed have a lower pO2 than vascular retina at the optic nerve head.52 However, no spatial or temporal associations between this presumed hypoxia and NV

incidence and severity were provided, so the role of hypoxia in NV could not be determined from these studies.

Macrophage infiltration has been linked to abnormal vessel growth in retinal NV models and in tumors.54,55 Thus, another possibility is that some combination of hypoxia and inflammation is involved in NV appearance. Studies using oxygen-induced retinopathy models have suggested the importance of macrophages and inflammatory factors in NV development.54,56 For example, angiogenic factors, such as tumor necrosis factor-alpha (TNF-alpha) and cyclooxygenase (COX-2), which are associated with inflammation, are also upregulated in a mouse oxygeninduced retinopathy model. In the branch retinal vein occlusion study discussed above, an inflammatory reaction to a laser burn could have been a confounding factor involved with NV growth.19

In models of diabetic retinopathy, it has been suggested that early activation of leukocytes leads to monocyte adhesion to the capillary endothelium (leukostasis) with subsequent decreases in retinal blood flow and, ultimately, retinal hypoxia.21,57 Linsenmeier et al. measured intraretinal oxygen profiles using electrodes in 3 long-term (> 6 years) diabetic cats and compared these data to prior results generated in their lab from control cats. They found evidence for retinal hypoxia, which appeared to be correlated with microaneurysms, leukocyte plugging of vessels, and/or endothelial cell death.21 Importantly, retinal NV was not found. This study clearly demonstrates that the link between hypoxia, NV, and inflammation is not well understood, since retinal hypoxia, which is found during chronic diabetes and expected to be linked with upregulated inflammatory factors, did not lead to NV.

To also investigate these issues, Handa et al. tested the possibility that hypoxia and inflammation co-exist before retinal NV.46 In a fibroblastinjected eye of non-diabetic rabbits, there is cellular proliferation in the vitreous, NV, and retinal detachment. Antoszyk et al. suggested that NV growth in this model is partly attributable to inflammatory mediators such as macrophages.58 Handa et al. used 19F NMR of a small perfluorocarbon (PFC) droplet placed in the vitreous on the surface of the retina. Significantly lower than normal preretinal vitreous oxygen tensions were found from the first day after cell injection until the development of visible NV, without coexisting evidence for vascular occlusion or retinal detachment. These data support the suggested notion that some combination of hypoxia and inflammation can combine to generate NV.

In what is perhaps the most comprehensive study to date, Zhang et al. measured preretinal vitreous pO2 using 19F NMR and a perfluorocarbon droplet in the newborn rat both during normal retinal vessel development and before and after appearance of retinal NV.49 The newborn rat model was

chosen for a number of reasons. Rats have no retinal circulation at birth (P0). The retinal circulation grows during P0–P14 via vasculogenesis and angiogenesis in a pattern similar to that described above for humans.59,60 For example, coverage of the retina by the superficial vessels at P7 in the rat is similar in appearance to that at 18-20 weeks post-conception in the human. However, while normal human retinal vessels develop in utero at systemic arterial oxygen tensions < 30 mm Hg, the rat retinal circulation develops primarily after birth at arterial oxygen levels of about 100 mm Hg. Despite these differences, Penn et al. have shown that alternating arterial oxygen levels in newborn rats between P0 and P14 produces blood oxygen levels similar to those found in at-risk infants and results in retinal NV after removal to room air.61 For example, daily alterations between 50% and 10% oxygen between days P0 and P14 followed by recovery in room air between P14 and P20 (the “50/10” condition) result in 100% of the rat pups having retinal NV (i.e., 100% incidence) and, when graded by the number of clockhours involved, 6 clockhour severity. To determine clockhour, an analog clock face is mentally superimposed on the retinal surface and the number of clock hours (a score from 0 to 12) occupied by abnormal vessel growth determined. To the best of our knowledge, there have not been full studies to determine whether or not there is are inflammatory components in this newborn rat NV model, although a proinflammatory isoform of VEGF (VEGF-164) appears upregulated and associated only with NV but not normal retinal vessel development.62,63

Zhang et al. found evidence for retinal hypoxia at the border of the vascular and avascular retina during normal retinal vessel growth (P1–P10), but not after the retina had fully vascularized (at times > P14).49 This demonstrates that hypoxia is normally involved in vasculogenesis (supporting the “physiological hypoxia” theory) and implies that lower than normal oxygen tensions do not necessarily result in abnormal retinal vessel growth. Zhang et al. also reported retinal hypoxia before the appearance of NV, but not after NV was evident.49 Importantly, although all of the border between vascular and avascular retina was likely hypoxic, NV was only found in about 50% of the hypoxic regions.51 This result is similar to that reported by Pournaras in a pig model in which NV occurred in less than 50% of the hypoxic regions (see above). Because the presence of hypoxia was not correlated with NV occurrence during normal and before abnormal vessel development, Zhang et al. concluded that hypoxia does not seem to cause the phenotype change from normal to abnormal vessel development.

In summary, when taken together, the above measurements of retinal oxygen levels strongly imply that hypoxia at the border of vascular and avascular retina is not a causative factor in the development of retinal NV.

6.RETINAL OXYGENATION RESPONSE: AN ALTERNATIVE HYPOTHESIS

6.1Rationale

Since it does not appear that hypoxia at the border of vascular and avascular retina per se is pathogenic (see above), perhaps changes in oxygenation ability may be linked with NV. There are clinically observable changes to the entire retinal circulation. For example, in ROP, retinal vessels posterior to the border between vascular and avascular retina can become dilated and tortuous (also known as plus disease), and this vascular abnormality has been linked with visual complications.64,65 In patients with diabetes, diffuse retinal edema (measured by leakage of fluorescein from extensive areas of posterior retinal capillary) has now been characterized and is associated with vision loss.66,67

How might panretinal oxygenation changes contribute to retinopathy? We first note that the retinovascular system is never at steady state and must constantly adapt. If the entire retinovascular system is unable to appropriately adjust, a panretinal dynamic mismatch between oxygen supply and demand can occur. This oxygen supply dysfunction can then increase the risk of retinopathy either developing or progressing. In other words, the presence of panretinal vascular abnormalities may prevent oxygen supply from adequately satisfying oxygen demand, not just during room air breathing but during conditions of normal retinovascular activity.

One approach to determining if there is a defect in the retinovascular system’s regulatory response would be to use a provocation test. Such a test could also be envisioned as the foundation for a clinical test that predicts the course of diabetic retinopathy and its response to treatment. This approach can potentially produce important insights into the pathophysiological basis of the disease and reveal novel targets for therapy.

6.2Method

We have developed a functional MRI method that detects a carbogeninduced increase in vitreous partial oxygen pressure over the room air value ( pO2) as an increase in the signal intensity.68-72 It is important to note that steady-state (room air) vitreous oxygen tension cannot be measured using MRI, because many factors (e.g., vitreous temperature and protein content) can unpredictably alter the baseline preretinal vitreous water signal and its relaxation properties. However, these factors are not likely to change on the short time scale between baseline and carbogen breathing. Thus, their