for AR (152), has made it possible to target specifically the role of the polyol pathway in the early stages of the development of experimental diabetic retinopathy. ARI-809 administered to diabetic rats at doses documented to inhibit both sorbitol and fructose accumulation in the retina, reproduced exactly all preventative effects on retinopathy (16, 19, 20) observed with the less specific (168) ARI sorbinil (140). On this basis, it can be stated that aldose reductase is itself the key relay that converts hyperglycemia into glucotoxicity for specific cell types in the retina. This conclusion is bolstered by the robust protection against the effects of diabetes observed in the retinas of 15-month-old db/db mice lacking the AR gene product (87).

As to the nature of the glucotoxicity generated through AR activity in the retinal vessels and leading to the characteristic histopathology of diabetic retinopathy, gene expression profiling points to the concurrence of multiple events, but identifies oxidative stress and proinflammatory changes as uniquely induced by excessive polyol pathway activity (169). This “signature” may become a useful reference when seeking evidence for polyol pathway activity in human diabetic retinopathy.

THE POLYOL PATHWAY IN HUMAN DIABETIC RETINOPATHY

The Sorbinil Trial

The Sorbinil Retinopathy Trial has been the only major clinical trial testing an AR inhibitor on diabetic retinopathy. In this multicenter, randomized, placebo-controlled, double-blind study, 497 patients with insulin-dependent diabetes and absent to mild retinopathy were followed for a median of 41 months. The sorbinil-treated (250mg/day) group was found not to differ from the placebo-treated group in terms of progression of retinopathy, although the number of microaneurysms increased at a slightly slower rate in the sorbinil-treated group (170). Knowledge gained since the trial warns that the findings are not readily interpretable. First, the efficacy of sorbinil was monitored by measuring in erythrocytes the levels of sorbitol, an imprecise indicator of flux, and now known to be a poor predictor of the functional benefits of ARIs (156). Moreover, erythrocyte sorbitol levels remained 26% above normal, and there was no information of an effect of sorbinil on the polyol pathway in retinal target cells. Although sorbinil could not have been used in larger doses on account of the risk of side effects in humans, the drug was given at a dose corresponding to 3.5mg/kg/day, almost 20-fold lower than the dose effective in prevention of retinopathy in diabetic rats (16). It is therefore probable that the dose of sorbinil used in the Sorbinil Trial was insufficient to silence the polyol pathway and did not permit testing the role of the pathway in diabetic retinopathy. Additionally, approximately half of the study population had some degree of retinopathy, and the treatment lasted a little over 3 years. We have since learned from the Diabetes Control and Complications Trial (DCCT) that in diabetic retinopathy prevention is much more effective than intervention, and that 3 years are grossly insufficient to demonstrate the efficacy even of treatments, such as improved glycemic control, that have a priori a high likelihood of success (171).

Evidence Supporting Polyol Pathway Activity and Functional

Importance in Human Diabetic Retinopathy

The negative results of the Sorbinil Trial could be falsely negative, or may instead reflect that the polyol pathway is not active and/or not pathogenic in human diabetes.

Only the appropriate tools, i.e., new ARIs with a therapeutic index higher than those of the older drugs, and clinical trials better controlled and better designed than the Sorbinil trial will finally, bring the polyol pathway hypothesis to rigorous testing in human diabetic retinopathy.

Several observations support a continuing interest in testing the hypothesis. Studies in postmortem human eyes have shown in retinas from diabetic patients with retinopathy more abundant AR immunoreactivity in ganglion cells, nerve fibers, and Müller cells, as compared to retinas from nondiabetic individuals (172). After negative findings by several investigators in the past, we have documented unequivocally that human retinal endothelial cells contain AR (16) (Fig. 6). Insofar as the good health of endothelial cells is critical to the highly regulated permeability and structural integrity of the retinal capillaries, excess glucose flux through the AR of retinal endothelial cells becomes a strong candidate mechanism for the disruption of barrier properties and the capillary obliteration characteristic of human diabetic retinopathy. We have reviewed above the presence of AR in other retinal cell types, such as pericytes and Müller cells, also affected in human diabetes. Finally, human retinas from nondiabetic eye donors accumulate sorbitol when exposed to high glucose in organ culture (16) (Fig. 7). The extent of accumulation is quite comparable to that occurring

Fig. 6. Aldose reductase in human retinal endothelial cells. Fresh retinas obtained from postmortem eyes of nondiabetic donors were incubated with collagenase type 1, and the dissociated cells were fixed briefly with acetone and immunostained with antibodies to AR and von Willebrand factor. (A) and (B) show AR immunoreactivity (green) in cells manifesting the granular perinuclear fluorescence of von Willebrand factor (red) characteristically seen in retinal endothelial cells in situ (179). The AR antibodies used in (A) were a gift from D. Carper, those used in (B) from R. Sorenson. Panel (C) shows cells from the same preparations staining only for AR or for neither protein. Bar = 20 m. Copyright © 2004 American Diabetes Association (from (16) reprinted with permission from the

American Diabetes Association).

Fig. 7. Aldose reductase activity in human retina. Fresh retinas obtained from postmortem eyes of nondiabetic donors were exposed for 24 h in organ culture to normal (5 mmol/l, blue bar) or high (30 mmol/l, red bar) glucose, and sorbitol levels were measured. Fresh retinas obtained from normal rats were tested in parallel. In (A), the bars represent the mean ± SD of the measurements performed in the indicated number of individuals. *P < 0.01 vs. normal glucose. (B) Presents a hemoglobin (Hb) immunoblot performed to assess the quantity of erythrocytes trapped in the blood vessels of the human retinas and potentially contributing to sorbitol accumulation. Protein lysate (20 g/lane) from fresh human retina or retina incubated in normal (N) or high (H) glucose was subjected to SDS–PAGE together with human Hb standards and probed with antibodies to Hb. Hb levels in the whole human retina did not exceed 40 g, whereas both the basal and stimulated levels of sorbitol were of the magnitude measured per gram Hb in human erythrocytes. This documented that resident cells of the human retina metabolize glucose to sorbitol when exposed to high glucose. Copyright © 2004 American Diabetes Association (from (16) reprinted with permission from the American Diabetes Association).

in the normal rat retina incubated in parallel, indicating that human retinal AR is readily responsive to hyperglycemia. The human enzyme is in fact widely used as a transgene in mice to confer susceptibility to diabetic complications, from cataract (100), to atherosclerosis (173). The evidence that the polyol pathway can be activated in the human retina in the presence of high glucose permits anticipation of tissue consequences, and complements in this respect the information from the human genetic studies reported earlier (see the sections “Polymorphisms of the AR Gene” and “AR Polymorphisms and Risk of Diabetic Retinopathy”) that alleles associated with elevated AR expression are also associated with accelerated development or progression of human diabetic retinopathy.