rate-limiting enzyme in the polyol pathway. AR inhibitors prevent in most cases the rise in tissue polyol and, at sufficiently high dose, tissue damage observed in experimental galactosemia (6). This does not mean that high glucose causes tissue damage via precisely the same mechanisms as high galactose, considering that glucose is substrate for several enzymatic and nonenzymatic reactions in addition to those of the polyol pathway, and is metabolized differently from galactose even in the polyol pathway. That is, galactose is metabolized by AR more avidly than glucose, and galactitol, the product of galactose metabolism through AR, is a very poor substrate for further metabolism by the second enzyme in the pathway, sorbitol dehydrogenase (7–9); therefore, galactitol accumulates to an even greater degree than sorbitol. [Although galactitol is a poor substrate for the second step of the polyol pathway, galactose can cause a rise in NADH/ NAD+ by another pathway, likely the galactonate pathway (10)]. In any event, the galactosemic model – where hyperhexosemia in the presence of normal insulin action is the trigger, and aldose reductase is a key transducer from hyperhexosemia to retinal histopathology that closely resembles human diabetic retinopathy (3, 11) – has provided an important and testable paradigm for the pathogenesis of diabetic retinopathy.

Indeed, several experimental findings over the years have made the polyol pathway an attractive mechanism for the characteristic lesions of human diabetic retinopathy. As will be described in more detail below, certain polymorphisms in the promoter region of the aldose reductase gene are associated with susceptibility to, or more rapid progression of, diabetic retinopathy (12). Polyol pathway activity can damage cells by multiple mechanisms ((13), and later in this chapter), and all cell types that in the human retina are affected by diabetes contain AR (14–16). The neuroglial abnormalities now known to be part of both human and experimental diabetic retinopathy (17, 18) are attributable in experimental animals to polyol pathway activity (19, 20), as is the spectrum of early and late vascular abnormalities (16). But the evidence necessary to implicate mechanistically the polyol pathway in the pathogenesis of human diabetic retinopathy is not yet in. In this chapter we aim to review in some detail the features of the polyol pathway, the reasons for its attractiveness as a mechanism and a target, and the steps that must be taken to prove or disprove that it is relevant to human diabetic retinopathy.

BIOCHEMISTRY AND GENETICS OF THE POLYOL PATHWAY

The polyol pathway consists of two soluble cytoplasmic enzymes, AR and sorbitol dehydrogenase (Fig. 1a and b, respectively). The first enzyme, AR, can convert intracellular glucose and NADPH to sorbitol and NADP+, while the second enzyme, sorbitol dehydrogenase, reversibly changes sorbitol and NAD+ into fructose and NADH. In the diabetic state, chronic hyperactivity of this pathway is driven by “hyperglysolia,” i.e., chronically elevated cytosolic glucose concentration, and underlies shifts in cellular redox, osmotic, and antioxidant systems.

Aldose Reductase

THE ALDOSE REDUCTASE ENZYME

Aldose reductase (E.C. 1.1.1.21; abbreviated ALD2, AKR1B1, or in this chapter, AR) was first described by Hers (21) and belongs to the aldo-keto reductase superfamily (22).

Fig. 1. X-Ray structures of the two enzymes of the polyol pathway, aldose reductase and sorbitol dehydrogenase. Shown in Fig. 1a is human aldose reductase (AR) with a bound molecule of nicotinamideadenine dinucleotide phosphate (NADPH) cofactor as well as a molecule of aldose reductase inhibitor (ARI) ARI-809. The asterisk (*) designates the C4 carbon of the nicotinamide ring of NADPH, the site of the substrate hydride transfer from the C4 carbon of the nicotinamide ring to the C1 aldehydic carbon of the substrate, e.g., glucose (not shown). Prepared by PJO from Protein Data Bank (http://www. rcsb.org) entry 1Z89 (177) using MoVit v. 2.0. Shown in Fig. 1b is a human sorbitol dehydrogenase (SDH) monomer of an SDH tetramer with a bound molecule of oxidized nicotinamide-adenine dinucleotide (NAD+) cofactor and a catalytic zinc atom (Zn) (sphere). The asterisk (*) designates the C4 carbon of the nicotinamide ring of NAD+, the site of hydride transfer between the nicotinamide ring and the C2 carbon of the substrate, e.g., sorbitol or fructose (not shown). Prepared by PJO from Protein Data Bank (http://www.rcsb.org) entry 1PL8 and from data in (50), using MoVit v. 2.0.

AR is a monomeric enzyme of 35,900Da with an (α/β)8 structure composed of eight beta-pleated sheet segments comprising a β-barrel surrounded by eight alpha-helices and two small accessory helices (Fig. 1a). AR is localized to cell cytoplasm and contains no metal ion or carbohydrate (23,24). The enzyme preferentially uses NADPH as a hydride donor to reduce aldehydic carbons to the corresponding alcohol, e.g., glucose to sorbitol.

Aldose reductase has broad substrate specificity, but its “natural” substrate in most tissues is unknown. Its kinetic constants are such that it functions primarily in the reductive direction, that is, to reduce aldehydes to corresponding alcohols (25). Its kinetic mechanism is formally categorized as compulsory ordered bi–bi with coenzyme binding first, then binding of aldehyde followed by alcohol leaving first and oxidized coenzyme NADP+ leaving last (26). Although the Km for glucose is often reported as 70–400 mM, AR acts catalytically only on the low abundance (0.0023%) aldehydic, straight-chain form of glucose, as opposed to the much more prevalent cyclic anomeric “boat” and “chair” forms of glucose (27, 28); it transforms the straight-chain form of glucose with a Km of 5 µM (29). AR transforms glucose into sorbitol relatively slowly with a kcat of30 min−1 (29). For further details on the structure and enzymology of AR, the reader is referred to recent studies and previous reviews, e.g., (26, 30, 31).

THE ALDOSE REDUCTASE GENE

Location and Structure of the AR Gene. The aldose reductase gene (AKR1B1) has been localized to human chromosome 7 locus q35 (32) and is distributed over 18 kilobases (kb) that contain ten exons coding for 316 amino acids (33, 34) (Fig. 2).