Polyol pathway activity appears thus sufficient to initiate the oxidative stress that is observed in the diabetic retina and to account for cellular consequences that have the potential to shape the development and progression of diabetic retinopathy. To prove or disprove a causative role of the polyol pathway in the lesions of diabetic retinopathy, studies have been undertaken over the years using AR inhibitors (ARIs). These drugs have been useful tools and have provided results that justify a continuing interest in the polyol pathway. However, the results also urge the development of new types of drugs and the testing of new hypotheses. We provide below highlights of studies with ARIs used to date that are instrumental to the well-informed interpretation of the results obtained in diabetic retinopathy.

ARI STRUCTURES AND PROPERTIES

ARIs have been reviewed on numerous occasions, and the reader is directed to the published reviews for details, e.g., (6, 96, 143–146). Selected ARIs of current interest or of particular historical interest to diabetic retinopathy are shown in Fig. 4. There have been three major chemical classes of ARIs (1) spiro-imides, exemplified in Fig. 4 by sorbinil, fidarestat, and ranirestat; (2) carboxylic acids, illustrated by tolrestat, zopolrestat, zenarestat, and epalrestat; and (3) pyridazinones, represented by ARI-809. Each class has had distinctive strengths and limitations in terms of in vitro potency, in vivo potency, pharmacokinetic properties, and safety. With few exceptions, toxicity has been unique for each ARI and apparently not related to AR inhibition.

For reasons of cost, speed, and presumed tissue sensitivity, most ARI clinical trials have been conducted against diabetic neuropathy. Epalrestat (Fig. 4) is commercially available (only) in Japan for this indication, and a multiyear study has recently reported a protective effect of chronic epalrestat use on diabetic nerve function (147), although the study was not double blinded (148). Fidarestat and ranirestat (Fig. 4) have been studied for improvement of diabetic neuropathy, but have given disappointing results in their Phase 2 (149) and Phase 3 (150) trials, respectively. However, Phase 3 study of ranirestat for prevention of progression of diabetic neuropathy and retinopathy is currently ongoing (150). A recent analysis based on translational pharmacology data in preclinical models of diabetic neuropathy, suggests that most ARIs have been used in human neuropathy trials at doses that were of inadequate functional potency against the endpoints tested (151). This appears to have resulted from relying on nerve sorbitol levels as the primary biomarker of tissue effect, which caused an 20to 40-fold overestimation of in vivo potency. Thus, despite early encouraging results, sorbinil, tolrestat, zopolrestat, and zenarestat (Fig. 4) are no longer in development. Nor is the newly discovered pyridazinone ARI-809 (152), recently found to absorb UV-A and UV-B and to cause exacerbation of spontaneous light-induced retinal damage in albino rats after 6 months of dosing (153). However, the discovery of the pyridazinone class demonstrated that it is possible to find new classes of ARIs with in vivo potencies high enough to exert robust in vivo antioxidant activity as opposed to simply lowering tissue sorbitol (151).

ARIs that penetrate lens have shown in most cases robust activity against retinopathy when given at relatively high doses in preclinical models (see later), but there has been only one major ARI trial for diabetic retinopathy, the Sorbinil Retinopathy Trial, which was unsuccessful and will be discussed further below.

Fig. 4. Structures of selected aldose reductase inhibitors (ARIs). See text for further discussion.

EFFECTS OF ARIS IN EXPERIMENTAL DIABETIC RETINOPATHY

It has become evident over the years that the species of the experimental animal model and the dose of ARI are important determinants of the effect on retinopathy. Whether induced by galactosemia or diabetes, retinopathy is more consistently susceptible to prevention by ARIs in the rat than in the dog model (reviewed in (154)). The reasons have not been investigated systematically, but may relate, at least in part, to the doses of ARIs used in the two species and relevant pharmacokinetics. In diabetic dogs, the ARI sorbinil used at a dose of 20 mg/kg/day failed to prevent the typical vascular lesions of retinopathy (155). This could not be attributed to absent activity of the polyol pathway in the retina of diabetic dogs, insofar as the retinal concentration of sorbitol increased by the approximately threefold observed in most models, and the increase was prevented by the dose of sorbinil used (155). Studies on the role of the polyol pathway in the neuropathy of diabetic rats made clear that inhibition of tissue fructose accumulation predicted the efficacy of ARIs on outcomes related to neuropathy better than inhibition of sorbitol accumulation (151, 156, 157). This is consistent with the hypothesis that tissue damage is a consequence of the flux of glucose through the pathway altering homeostasis at several levels (see earlier), rather than of the amount of tissue sorbitol measurable at any given time (29,102). The reason normalization of abnormal metabolic flux through the polyol pathway – as opposed to normalization of tissue sorbitol – is

essential in order to prevent tissue damage is because the abnormal flux is closely linked to the generation of oxidative stress (Fig. 3) (151, 158).

When tested on functional abnormalities of the retina or retinal vessels in diabetic rats, sorbinil at a dose of 10 mg/kg/day reduced but did not prevent deterioration of the electroretinogram (159), and different ARIs prevented in a dose-related manner albumin permeation (160). Among the characteristic structural changes of retinal vessels, basement membrane thickening was only partially prevented in diabetic rats by ponalrestat in experiments that did not however document the effect of the dose used on polyol pathway metabolites (161); fidarestat prevented basement membrane thickening in a dose-related manner, while pericyte loss appeared sensitive to all doses of fidarestat tested (162).

Studies performed more recently in the diabetic rat have taken such dose considerations into account and used doses of ARIs documented in advance to prevent or decrease substantially retinal fructose accumulation. It is of note that for both sorbinil and ARI809, two structurally different ARIs, the doses used to normalize retinal fructose in diabetic rats (65 and 50 mg/kg/day, respectively) led to lowering of sorbitol below control levels (19, 140). This is consistent with the pattern seen with functionally efficacious doses in diabetic rat nerve, e.g., (156, 157), and confirms that doses targeting fructose levels inhibit the pathway to a greater extent than doses targeting, and often not even normalizing, sorbitol levels.

The streptozocin diabetic rat is currently the most comprehensive model for human diabetic retinopathy. Like diabetic humans, diabetic rats show in the retinal capillaries apoptosis of pericytes and endothelial cells (141), deposition of complement (163), and ultimately development of pericyte ghosts and acellular capillaries (16, 141). Diabetic rats also show the Müller cell reactivity (18) and apoptosis of neurons in the ganglion cell layer (142) observed in human diabetes. All these abnormalities are prevented in the diabetic rat by different ARIs administered at doses that inhibit retinal polyol pathway activity (16, 19, 140).

The recent results in diabetic rats are not only proving a role for the polyol pathway in the spectrum of vascular, glial, and neural abnormalities that diabetes induces in the retina; they are also showing that ARIs are, to date, the only drugs that can prevent the whole spectrum of abnormalities (Fig. 5). When sorbinil and aspirin were compared to each other in the same experiments, both prevented the development of acellular capillaries, but only sorbinil prevented neuronal apoptosis and Müller cell reactivity (20). This could have translational importance if we were to learn that the early neural and/ or glial abnormalities occurring in the diabetic retina impact on clinically important aspects of retinopathy. For example, proper Müller cell function may be critical to the prevention and/or resolution of macular edema (164), and there is evidence that diabetes affects the regulation of water channels in these cells (165).

New observations are also defining with increasing precision the mechanisms whereby the polyol pathway causes the retinal abnormalities. ARIs inhibit AR preferentially, and can inhibit also aldehyde reductase, another enzyme in the aldo–keto reductase superfamily that plays a role in the detoxification of reactive aldehydes. This has generated the question of whether inhibition of aldehyde reductase (166) contributes to the beneficial or to the unwanted effects of ARIs, especially at the higher doses (167). Availability of ARI-809 (Fig. 4), characterized as one of the most potent and selective ARIs yet described with an IC50 for aldehyde reductase of 930 nM as compared to 1 nM

Fig. 5. Comparison of the effects of clopidogrel, aspirin, and ARI sorbinil on neuronal apoptosis and glial reactivity in the retina of rats with 2.5 months of streptozotocin-induced diabetes. In these experiments, rats were randomized from the time of diabetes induction to treatment with clopidogrel (Clop) (10 mg/kg/day), used as a selective antiplatelet agent, aspirin (ASA) (30 mg/kg/day) used as anti-inflammatory and antiplatelet agent, and ARI sorbinil (Sorb) (65 mg/kg/day) to test the effects of the three drugs on early and late vascular, neuronal, and glial abnormalities caused by diabetes in the retina. Clopidogrel had no effect on any abnormality tested, aspirin prevented all capillary abnormalities (results not shown in this figure), but only sorbinil prevented both the capillary and the neuroglial abnormalities. Neuronal apoptosis (A) was measured by counting TUNEL-positive nonvascular nuclei in whole retinas mounted and observed vitreal side up. In the boxplots, the bars encompass from the 90th to the 10th percentile of the scores and the box from the 75th to the 25th percentile; arrows point to the median. Glial reactivity was assessed by observing (B) the pattern of glial fibrillary acidic protein (GFAP) immunostaining in retinal sections (diabetes causes GFAP to be prominently expressed in the processes of the Müller glial cells that span the thickness of the retina; GCL ganglion cell layer; INL inner nuclear layer; ONL outer nuclear layer), and by measuring the levels of retinal GFAP by immunoblot ( (C) shows a representative immunoblot, and (D) the quantitation of the signals from immunoblots of retinal GFAP). C control rats, D diabetic rats. *P < 0.006 vs. control rats; **P < 0.002 vs. diabetic rats. Copyright © 2005 American Diabetes Association (from (20) reprinted with permission from the American Diabetes Association).