Is diabetic retinopathy an inflammatory disease?

L ast year an ARVO symposium organised by Tim Kern was titled “Is diabetic retinopathy an inflammatory disease?” It is a timely question. chronic subclinical inflammation may underlie much of the vascular pathology of diabetic retinopathy. Nomenclature is critical to this discussion, so the definition of inflammation bears repeating here. Macroscopic inflammation comprises the classic signs of pain (dolor), heat (calor), redness (rubor), swelling (tumor), and loss of function (functio laesa).¹ None of these signs, except for swelling and loss of function, clearly applies to diabetic retina. However, at a microscopic level, inflammation consists of vessel dilatation, altered flow, exudation of fluids, including plasma proteins, and leucocyte accumulation and migration.¹ Given the recent data that have been generated in relevant models of diabetic retinopathy, the latter definition appears to fit. A brief overview of the evidence illustrates this point.

Within one week of experimental diabetes, leucocytes adhere to and accumulate within the vasculature of the retina.^2,3 A subset of these leucocytes exit the vasculature and transmigrate into the neural retina.^2,4 The leucocyte increases are moderate in nature and precede any overt clinical evidence of retinopathy. However, they progress with time.² Monocytes and neutrophils predominate,⁴ although preliminary data suggest that lymphocytes may be involved as well (Ahmed, Ishida, Adamis, et al, unpublished observation). The leucocytes actively tether themselves to the endothelial cell lining via classic adhesion molecules, including intercellular adhesion molecule-1 (ICAM-1) on the vasculature² and β₂ integrins on the leucocytes.³ The expression levels for these adhesion molecules increase in early diabetes and correlate with the leucocyte increases.^2,3 Additional adhesion molecules, including vascular cell adhesion molecule-1 (VCAM-1) and VLA-4 may also be involved, but here the data are more preliminary.⁵

The leucocyte increases coincide with the onset of diabetic vascular dysfunction. At first, the dysfunction is subclinical in nature, probably because of the lack of sensitivity of our current clinical detection methods. However, when more powerful experimental techniques are applied, the early alterations uncovered include a subtle breakdown of the blood-retinal barrier, premature endothelial cell injury and death, and capillary ischaemia/reperfusion.^2,6–8

The leucocytes appear to be causal for these pathologies. When diabetic rats are treated with ICAM-1 or β₂ integrin neutralising antibodies, leucocyte adhesion is suppressed,^2,3 blood-retinal barrier breakdown is normalised,² and endothelial cell injury and death are prevented.⁸ When mice deficient in the ICAM-1 or β₂ integrin gene CD18 are made diabetic and followed for 11 months, the retinal vasculature is indistinguishable from age matched normal non-diabetic mice (Joussen, Poulaki, Adamis, et al, unpublished data). In contrast, the retinas from diabetic ICAM-1 and CD18 competent mice exhibit marked increases in leucocyte density, blood-retinal barrier breakdown, and endothelial cell injury and death. Since 11 months represents almost half of the normal mouse life span, these data suggest that the inhibition of leucocyte adhesion provides effective long term suppression of certain diabetes related pathologies.

Inflammation may represent the inciting and final common pathway leading to the complex pathology that is diabetic retinopathy

If diabetic retinopathy is an inflammatory disease, then one would expect that anti-inflammatory drugs would have a beneficial effect. This seems to be the case. In 1964, it was observed that patients with rheumatoid arthritis receiving high doses of aspirin tended to have less severe diabetic retinopathy.⁹ Recently, Kern and Engerman directly tested the effect of aspirin in the relevant dog model of diabetic retinopathy. When started shortly after the onset of diabetes and given for 5 years, aspirin prevented certain classic histopathological features of diabetic retinopathy.¹⁰ Acellular capillary formation, retinal haemorrhage development, and capillary sudanophilia (a non-specific indicator of cellular degeneration) were all inhibited. A trend towards microaneurysm and pericyte ghost suppression was also observed, although statistical significance was not achieved. The moderately high doses utilised (20–25 mg/kg) were anti-inflammatory in nature and more than twice the antiplatelet dose utilised in the negative Early Treatment Diabetic Retinopathy Study aspirin trial (650 mg/day).¹¹ Recent data in a rodent model complement and extend these findings by showing that a variety of anti-inflammatory drugs can be effective.¹² High dose aspirin (50 mg/kg), meloxicam (cyclo-oxygenase-2 inhibitor), and etanercept (soluble tumour necrosis factor α receptor) each potently suppressed diabetic retinal ICAM-1 expression, leucocyte adhesion, and blood-retinal barrier breakdown. Anti-inflammatory drug testing in human diabetic retinopathy has begun. Envision, a sustained release formulation of fluocinolone implanted into the vitreous, appears to resolve refractory diabetic macular oedema in early clinical testing (Andrew Pearson, MD, Foundation Fighting Blindness Drug Delivery Meeting, San Francisco, 2001).

How is the slowly evolving nature of diabetic retinopathy reconciled with the finding of early onset inflammation? As noted above, the inflammation is, at first, mild and subclinical in nature. It does not result in an overt vasculitis. However, because the inflammation is chronic in nature, the damage to the vascular endothelium is cumulative. The endothelial cells of the diabetic retinal vasculature proliferate and die at rates much higher than normal.^13,14 The vascular injury, in large part, is leucocyte induced.⁸ As first proposed by Mizutani and coworkers,¹⁴ we hypothesise that the chronic low grade endothelial injury of early diabetes is reparable, but as diabetes progresses, the vascular endothelium reaches its Hayflick number and can no longer proliferate and repair the damaged endothelial lining. Acellular capillary formation ensues at this point. We speculate that with the formation of acellular capillaries, irreversible ischaemia develops, leading to marked retinal VEGF upregulation and transition to the proliferative stage of retinopathy.

How is VEGF, a molecule causally linked to the pathogenesis of diabetic retinopathy, connected to the inflammation? It is known that retinal VEGF expression is correlated with diabetic blood-retinal barrier breakdown^15–17 and ischaemia related neovascularisation in animals^18,19 and humans.^20–22 Moreover, the inhibition of VEGF prevents these processes in relevant experimental models.^17,19,23 New data now indicate that VEGF can also trigger inflammation.^24–27 In the retina and elsewhere, VEGF can induce ICAM-1 expression and leucocyte adhesion.^24–27 Further, monocytes, via VEGF receptors on their surfaces, migrate in response to VEGF.²⁸ When retinal ICAM-1 bioactivity is blocked, VEGF induced blood-retinal barrier breakdown is almost completely prevented, demonstrating the mechanistic link between the permeability and inflammation enhancing effects of VEGF.²⁶ In experimental diabetes, retinal VEGF levels increase within one week. When the endogenous retinal VEGF bioactivity is blocked with a soluble receptor, the ICAM-1 upregulation, leucocyte adhesion, and blood-retinal barrier breakdown are all prevented.^17,29 Taken together, these data strongly suggest that retinal VEGF upregulation occurs early in diabetes and serves as an important upstream inducer of early retinal inflammation. What triggers VEGF expression in the first place? There are no definitive answers as yet. However, changes in glucose concentration per se can alter VEGF expression³⁰ and may represent a direct proximal upstream stimulus. As diabetes progresses, other stimuli, including advanced glycation end products³¹ and reactive oxygen intermediates,³² probably serve to further increase VEGF expression.

How relevant are the experimental data to human diabetic retinopathy? Rodent retinas, although lacking the ability to develop bone fide proliferative diabetic retinopathy, exhibit almost all the biochemical, pathophysiological, and histopathological features of background retinopathy. These include blood-retinal barrier breakdown, altered blood flow, vessel dilatation, microaneurym formation, basement membrane thickening, intraretinal microvascular abnormalities (IRMA), accelerated endothelial cell proliferation and death, pericyte loss, neural cell death, acellular capillary formation, haemorrhage formation, platelet microthrombi, and VEGF upregulation. However, care must to taken when extrapolating data generated in rodents to humans. An ongoing anti-ICAM-1 trial in Crohn's disease is showing some promise in humans. However, anti-CD18 trials for stroke and myocardial infarction, although supported by rodent data, failed quite spectacularly in humans. Yet the body of correlative evidence in the case of human diabetic retinopathy is greater than that for stroke and myocardial infarction. McLeod and coworkers, in a landmark early study, found a marked increase in leucocyte density and retinal vascular ICAM-1 immunoreactivity in human eyes with diabetic retinopathy.³³ Others have shown that human diabetic leucocytes are more activated and less deformable,³⁴ and that patients with diabetic retinopathy have a more pronounced stimulus induced expression of leucocyte adhesion molecules.³³ In addition, circulating adhesion molecule levels, shed from activated leucocytes and endothelium, are elevated in patients with progressively worsening retinopathy.³⁵ Nevertheless, more evidence is needed.

Taken together, a body of data has been generated indicating that diabetic retinopathy is a low grade inflammatory disease. Inflammation, specifically leucocyte adhesion to the retinal vasculature, may represent the inciting and final common pathway leading to the complex pathology that is diabetic retinopathy. However, more work remains to directly prove this hypothesis, especially in human diabetic retinopathy. The latter, after all, is the only model that really counts.