individuals with up to 13 different alleles in some populations. The most common allele in all populations is n = 24, called “Z.” An allele with two fewer bases, i.e., one less (AC) n unit, viz., n = 23, is designated “Z−2.” Similarly, a microsatellite allele with one additional AC unit compared to the most frequent, viz., n = 25, is referred to as “Z + 2,” etc. A second polymorphism of the AR gene is a C(−106)T single nucleotide polymorphism (SNP) in the basal promoter region (42). A third polymorphism is a BamHI site consisting of a single A to C substitution at the 95th nucleotide of intron 8 (43, 44) [IVS8A(+95) C in Fig. 2; also designated A(+11842)C (12)]. The (AC)n and C(−106)T polymorphisms are closely linked (42, 45, 46). A C(−12)G SNP has also been described (47).

Sorbitol Dehydrogenase

THE SORBITOL DEHYDROGENASE ENZYME

Sorbitol dehydrogenase (SDH) (48) (EC 1.1.1.14) belongs to a superfamily of medium-chain dehydrogenase/reductases (49), and the X-ray structure of human and rat SDH has been determined (Fig. 1b) (50, 51). The native enzyme is 140,000–160,000 Da and has four identical subunits of 354–356 amino acids that each contain one catalytic zinc and one NAD(H) binding site (29, 52, 53) (Fig. 1b). The N-terminal amino acid of human and sheep SDH is acetylated, e.g., (54). Interestingly, the tip of the “tail” of the SDH monomer (Fig. 1b) is predominantly hydrophobic, suggesting it could interact with a lipid environment. Although the subcellular localization of SDH is primarily cytosolic (55, 56), some of the SDH in human liver is found to be associated with the microsomal fraction (57) and multiple SDH isoforms have been reported (9, 58).

Each independent monomer of SDH stereospecifically oxidizes a spectrum of secondary alcohols (59, 60), including sorbitol which it reversibly oxidizes to fructose using coenzyme NAD+ (60). SDH likely binds and releases the straight-chain form of fructose (61); this conformer exists at 0.8% of the total ketose (62). Galactitol is metabolized weakly or negligibly by SDH, e.g., (7, 9, 63). Kinetic analysis of the mechanism of SDH reveals that it has a compulsory ordered reaction that is classified as Theorell-Chance bi–bi with coenzyme binding first and leaving last (64–66). The Km of

SDH for sorbitol is 1–4mM and its kcat is 100s−1 at pH 7.1 (29,63, 67, 68). Thus, the catalytic rate constant, kcat, of SDH is approximately three times higher than for AR.

THE SORBITOL DEHYDROGENASE GENE

Location and Structure of the SDH Gene. The SDH gene, SORD, resides on human chromosome 15 (69). Its position on this chromosome is reported at 15q15 (70, 71) or 15q21.1 (72). The human SORD gene has nine exons and eight introns and extends approximately 30 kb (73). Three Sp1 sites (CCCGCCCC) and a CACCC box were found in the 5′ noncoding region, but classical TATAA or CCAAT elements were absent, although a unique repetitive (CAAA)5 sequence was observed. In all tissues analyzed, two transcriptional initiation signals occur at 16 and 89 bp upstream of the translation initiation site for SDH. As seen for rat SDH mRNA (74), human SDH mRNA has an open reading frame that codes for 356 amino acids (73). In addition, a second ATG translation start site codon 126 bp upstream from the first start site was detected in sequencing rat testis SDH cDNA; in principle this could code for an

additional 42 amino acid N-terminal peptides in a pre-SDH (75). However, this peptide is probably removed in post-translational processing since pre-SDH has not yet been detected experimentally.

LEVEL OF EXPRESSION OF POLYOL PATHWAY ENZYMES

AND DIABETIC RETINOPATHY

Natural Variations in Polyol Pathway Enzyme Levels

and Diabetic Retinopathy

AR POLYMORPHISMS AND RISK OF DIABETIC RETINOPATHY

Certain polymorphisms of the AR gene have been linked in numerous, but not all, genotypic studies to faster or slower rates of development of diabetic retinopathy and other diabetic complications, e.g., (12, 76). In particular, the “Z−2” (AC)n microsatellite polymorphism, i.e., (AC)23 (Fig. 2), was originally discovered in association with rapid progression of diabetic retinopathy; i.e., Z−2 occurred in higher than expected frequencies in patients with type 2 diabetes who had retinopathy after a relatively short ( 5 years) duration of known diabetes (41). Subsequent studies in both type 1 and type 2 diabetic patients across different ethnic groups have found a positive association of Z−2, C(−106) T, and A(+11842)C polymorphisms of the AR gene (Fig. 2) with diabetic retinopathy (12, 77). One report found an association between C(−12)G, elevated AR transcription rate, and diabetic retinopathy (47). Negative studies may be attributable to differences in the sample size, patients’ characteristics such as duration of disease and genetic complexities in certain populations. For example, the CC genotype of the C(−106)T polymorphism, which caused higher AR transcriptional rates in vitro (78), was found associated with an approximately twofold increased risk of having proliferative retinopathy independent of other risk factors in Caucasian Brazilians, but not in African Brazilians (79).

The Z−2 allele of the AR gene was also found to be associated with approximately twofold higher levels of AR mRNA in peripheral blood monocytes of diabetic patients with nephropathy vs. diabetic patients without nephropathy or nephropathic patients without diabetes (80). In studies in vitro, constructs containing the Z−2 variant of the AR microsatellite resulted in rates of AR gene transcription 1.6- to 6-fold higher than constructs containing the Z + 2 or other variants of the microsatellite (78). In Japanese patients with type 2 diabetes, the Z−4, not the Z−2, polymorphism was found associated with proliferative retinopathy and with higher AR protein levels in the erythrocytes; while the Z + 2 allele was associated with absence of diabetic retinopathy (81). Likewise, the prevalence of diabetic retinopathy was observed to increase significantly with elevated erythrocyte AR levels in type 2 diabetic patients with duration of known diabetes of less than 10 years, but not with longer durations (82), another suggestion that increased AR expression may work to accelerate the development of retinopathy.

Conversely, several reports have linked polymorphisms associated with low expression of AR with lower than average frequencies of diabetic retinopathy, e.g., (81). Consistent with these and the aforementioned data, Zou and coworkers reported a twofold higher AR activity level in erythrocytes from Z−2/Z−2 patients than in erythrocytes from Z + 2/Z + 2 patients, although confirmation of the specificity of the enzyme activity assay (vs. aldehyde reductase activity) is warranted (83).

SDH POLYMORPHISMS AND DIABETIC RETINOPATHY

Although variations in the human SDH gene sequence have been detected, the impact of such variations on the expression of the gene or the prevalence and course of diabetic complications has not been determined (70).

Experimental Manipulations of Polyol Pathway Enzyme Levels

and Retinopathy

AR OVEREXPRESSION

Overexpression of human AR in diabetic mice accelerated diabetic neuropathy as manifested by a significantly increased drop in nerve conduction velocity and increased severity of nerve fiber atrophy in diabetic transgenic mice compared to nontransgenic diabetic littermates (84). However, no data for retinal endpoints in these AR transgenic diabetic mice are yet available. Another set of transgenic mice carrying human AR was studied after receiving for only 5–7 days diets high in glucose or galactose; in mice fed a diet containing 20% galactose for 7 days, the ocular pathology observed was cataract and occlusion of the retinal-choroidal vessels (85). However, in these transgenic mice no data for retinal endpoints were reported at later times or at any time in which diabetes was also present.

SDH OVEREXPRESSION

Bovine retinal capillary pericytes that were exposed to 30 mM glucose had modestly increased reactive oxygen species (ROS) generation, reduced DNA synthesis, and upregulated VEGF expression; under the same conditions, SDH overexpression significantly stimulated ROS generation and accentuated the cytopathic effects of glucose in an ARIand antioxidant sensitive manner (86). These data strongly suggest that elevated metabolic flux through the SDH step of the polyol pathway, as well as through the AR step of the polyol pathway, can contribute to ROS generation in retinal cells exposed to high glucose levels.

Transgenic mice that overexpress SDH have not been described to date.

AR “KNOCKOUT” MICE

Signs of diabetic retinopathy that include blood-retinal barrier breakdown, loss of pericytes, neuroretinal apoptosis, glial activation, and proliferation of blood vessels, were observed in 15-month-old db/db mice, and were all attenuated or prevented in db/ db mice with an AR null mutation (AR−/− db/db) (87). In the same study, AR deficiency also prevented diabetes-induced increased retinal nitrotyrosine staining, a marker of oxidative-nitrosative stress, reduction of platelet/endothelial cell adhesion molecule-1 expression, and increased expression of vascular endothelial growth factor, suggesting that AR is responsible for this spectrum of early events in the pathogenesis of diabetic retinopathy. The same group recently reported that AR deficiency prevented neuroretinal damage and glial activation induced by carotid artery transient ischemia (88), consistent with similar findings in cardiac tissue (89).

SDH-DEFICIENT MICE

No retinal endpoint data have been reported for C57BL/LiA SDH-deficient strain of mice that can be rendered diabetic with streptozotocin treatment (90).

MECHANISMS OF CELLULAR TOXICITY OF THE POLYOL PATHWAY

AND RELEVANCE TO DIABETIC RETINOPATHY

To begin entertaining a connection of the polyol pathway with diabetic retinopathy, it must be known that some critical cell types in the retina contain the enzymes of the pathway. This condition is well satisfied for AR, which is present in the vascular pericytes, endothelial cells, ganglion cells, and Müller glial cells of all species studied, including human (14–16, 91–93).

Chronic polyol pathway hyperactivity can impose on cells a variety of stresses, notably osmotic and oxidative stress, increased protein kinase C activation, and enhanced glycation via fructose and its metabolites leading to formation of advanced glycation endproducts (AGEs) (Fig. 3). Activation of the polyol pathway is also tightly coupled to activation of the pentose phosphate pathway (PPP) (7) which produces, among other metabolites, NADPH and glyceraldehyde-3-phosphate, the latter also strongly implicated in AGE formation (94, 95) (Fig. 3). Eventually, it will be important to know which of these stresses are operative in the individual retinal cell types that contain AR and undergo damage or death in diabetes. This knowledge may have therapeutic implications related to effective doses of AR inhibitors and identification of alternative drugs. For the moment, however, individual retinal cell types are not accessible with the rapidity required for direct biochemical and metabolic studies. We thus illustrate the biochemical and metabolic consequences of polyol pathway activation using data obtained in the whole retina and, mostly, in other tissues.

Osmotic Stress

AR reduces cytosolic glucose to sorbitol using NADPH as a cofactor. Sorbitol is an alcohol, polyhydroxylated and strongly hydrophilic, and therefore does not diffuse readily through cell membranes and can accumulate intracellularly with possible osmotic consequences (96). Of note, production of intracellular osmolytes to counterbalance extracellular hypertonicity is a physiological role of AR in the kidney medulla (97). Insofar as accumulation of 1 mol of membrane-impermeant solute per gram of intracellular tissue water will increase osmotic pressure by 1 mOsm per liter, elevation of intracellular sorbitol will trigger osmotic regulatory mechanisms (98). When such mechanisms, relatively unexplored in the retina, fail to fully compensate for increased intracellular sorbitol in the diabetic state, osmotic stress will result (Fig. 3). Probably only tissues and organs that accumulate concentrations of sorbitol in excess of 5 mol per gram will suffer osmotic consequences (99, 100). The increase in sorbitol concentrations measured in the whole retina of diabetic rats is not in the range that would generate osmotic stress (19, 101), but measurements performed in the whole organ may not be informative of events in discrete cell types. For example, a cell type that had an especially high ratio of AR to SDH could accumulate sorbitol to the point of generating intracellular hypertonicity, and yet the amount of sorbitol would be diluted substantially if the measurement, performed in the whole retina, included cell types not accumulating sorbitol. Additional studies are required to ascertain the susceptibility of individual retinal cell types to polyol pathway-induced osmotic stress.