Studies of Rat and Human Retinas Predict a Role for the Polyol Pathway in Human Diabetic Retinopathy Free

Experiments were performed following institutional guidelines for the use and care of laboratory animals. Sprague-Dawley male rats were randomly assigned to one of the following groups: control, diabetic, and diabetic treated with sorbinil. The experiments carried out for 2.5 months also included a group of control rats treated with sorbinil. Diabetes was induced with streptozotocin (57.5 mg/kg body wt) dissolved in citrate buffer, pH 4.5, and injected via the tail vein. Development of diabetes (blood glucose >250 mg/dl) was verified 3 days after the streptozotocin injection. Body weight was recorded three times a week in the diabetic rats, and 2–4 units of NPH insulin were administered subcutaneously as needed to prevent weight loss without preventing hyperglycemia. Sorbinil (Pfizer, Groton, CT) was given from the onset of diabetes at a dose of 65 mg · kg body wt⁻¹ · day⁻¹ mixed with the powdered diet. This sorbinil dose prevents the accumulation of both sorbitol and fructose in the retina of rats with 2.5 months of diabetes (12), indicating complete suppression of polyol pathway activity (4). Such preventative effects of sorbinil (given at 50 mg · kg⁻¹ · day⁻¹) have also been documented in rats with 10 months of diabetes (J.R. Williamson, personal communication). The rats were killed after 2.5 or 9 months of diabetes together with age-matched controls. At the time of death, blood was obtained by cardiac puncture for the measurement of GHb (Glyc-Affin GHb assay; PerkinElmer, Norton, OH). Whole eyes were embedded in OCT (optimal cutting temperature) compound on dry ice for sectioning or fixed in formalin for the preparation of retinal trypsin digests. The retinas that were immediately dissected were used for protein studies, for preparation of fresh microvessels and dispersed cells, and for organ culture experiments.

Human postmortem eyes were obtained from certified eye banks through the National Disease Research Interchange. The donors were 63, 66, and 75 years of age, two men and one woman, and all Caucasians. The cause of death was cancer for two donors and cardiac arrest for the third; none had diabetes. The time elapsed from death to enucleation was 4.6 ± 1.5 h and 30 ± 3 h to receipt of the eyes in the laboratory. The retinas were immediately isolated and placed in organ culture or used to prepare dissociated cells and protein lysates.

Retinas from rats with 9 months of diabetes and age-matched controls were fixed in 10% buffered formalin for 2–8 days and processed for trypsin digestion to isolate the intact microvascular network (20,24).

The intact microvascular network was isolated from fresh retinas of normal rats by the hypotonic lysis method as used previously (25).

Fresh rat and human retinas were incubated with 0.1% collagenase type I (Sigma, St. Louis, MO) in serum-free Dulbecco’s modified Eagle’s medium (DMEM) at 37°C for 50 and 70 min, respectively (26). The reaction was stopped by the addition of DMEM containing 10% FCS. Suspended cells were centrifuged at 200g for 5 min at 4°C, resuspended in PBS, and smeared on Superfrost/Plus Microscope Slides (Fisher Scientific, Pittsburgh, PA). The slides were air dried and stored at −20°C.

Retinas dissected sterilely from postmortem human eyes were cut into halves or quarters and placed flat onto the insert of a 6-well Transwell plate (Corning, Corning, NY). The tissue was covered with DMEM containing 10% FCS and antibiotics/antimycotics and placed in a tissue culture incubator at 37°C. The medium contained either 5 or 30 mmol/l glucose. Rat retinas were studied in parallel; each whole rat retina was flattened by two or three peripheral incisions and placed in organ culture onto the insert of a 12-well Transwell plate.

Retinal cryosections (6 μm) and dissociated cells were fixed in ice-cold acetone for 10 min; microvessels were fixed in formalin for 15 min. Immunohistochemistry was performed as described (12,25). The primary antibodies were mouse monoclonal antibody (mAb) 2A1 anti-rat C5b-9, to detect the membrane attack complex (MAC) of complement in rat retina (1:200; provided by W.G. Couser); mouse mAb anti-rat platelet endothelial cell adhesion molecule-1 (PECAM-1) (1:1,000; BD Pharmingen, San Diego, CA); mouse mAb anti-human von Willebrand factor (vWf) and rabbit anti-human vWf (both used at 1:500; Dako, Carpinteria, CA); and two different rabbit anti-rat aldose reductase (one provided by D. Carper and one by R. Sorenson, both diluted 1:1,000). Negative controls were obtained by substituting the primary antibodies with an equivalent concentration of nonimmune isotypic mouse or rabbit IgG as appropriate.

Semiquantitative scoring of MAC staining in rat retinal sections was performed by masked observers, taking into account the frequency of staining (number of retinal vessels positive for MAC over the total number of vessels detected in each of several fields by the vWf staining) and the area stained in each vessel (varying from the whole vessel wall to speckles). The scoring was from 0 to 4, with 0 indicating no staining and 4 uniform staining of all vessels.

Proteins were isolated from retinas and retinal microvessels lysed in radioimmunoprecipitation assay (RIPA) buffer (50 mmol/l Tris, pH 7.4, 150 mmol/l NaCl, 5 mmol/l EDTA, 1% Triton X-100, 0.5% Na deoxycholate, and 0.1% SDS) containing protease and phosphatase inhibitors (27). Protein concentration was determined with the bicinchoninic acid method, using BSA as standard (Micro BCA Protein Assay Kit; Pierce, Rockford, IL). Retinal proteins were resolved by SDS-PAGE and immunoblotted as previously described (27). Nonreducing conditions were used only for the study of CD59, which was detected with mouse mAb 6D1 (1:1,000; provided by B.P. Morgan [28]). CD55 (H-319), CD46 (H-294), and caveolin-1 (N-20) were detected with rabbit antibodies (all at 1:1,000; Santa Cruz, Santa Cruz, CA), occludin with mouse mAb anti-human occludin (1:1,000; Zymed, South San Francisco, CA), and neuron-specific enolase (NSE) with rabbit anti-cow NSE (1:20,000; Chemicon International, Temecula, CA). The blots were subsequently reacted with mouse mAb AC-15 anti–β-actin (1:200,000; Sigma) to verify even loading. Protein lysates from human retinas were tested for Hb to determine erythrocyte content using goat antibodies to human Hb (1:1,000; Bethyl Laboratories, Montgomery, TX). The human Hb used as standard was purchased from Sigma.

Rat retinal trypsin digests were studied with the In Situ Cell Death Detection Kit, Fluorescein (Roche Applied Science, Mannheim, Germany). The count of transferase-mediated dUPT nick-end labeling (TUNEL)-positive nuclei and their attribution to pericytes or endothelial cells were performed by masked observers as described previously (20). TUNEL-positive chromatin with uninformative shape or topography was counted as “undetermined.” Counts of TUNEL-positive cells are presented per whole retina.

Cell nuclei and acellular capillaries were counted in rat retinal trypsin digests stained with periodic acid Schiff hematoxylin and viewed at 20× magnification. Counts were performed in six to eight fields in mid-retina using the AxioVision4 program (Carl Zeiss, Göttingen, Germany). Nuclei were attributed to pericytes or endothelial cells on the basis of shape, location along the vessel, and intensity of staining. Pericyte nuclei are round or slightly oval and appear to protrude from the vessel wall, staining heavily with hematoxylin. Endothelial cell nuclei are ellipsoid in shape, larger, and lie in the axis of the capillary, taking a lighter stain. Capillary length was determined in each field, and counts are expressed as the number of cells per millimeter of capillary length. Acellular capillaries were those that had 1) no visible nuclei along the vessel between bifurcations, 2) length of at least 40 μm, and 3) width equal to at least 20% of an average capillary. Acellular capillaries counts are reported as number of acellular capillaries per mm² retinal trypsin digest.

The concentration of retinal sorbitol was determined by a modification (29) of the enzymatic method of Malone et al. (30). In the modified assay, after sorbitol is oxidized to fructose by sorbitol dehydrogenase and NAD⁺ concomitantly reduced to NADH, NADH is oxidized back to NAD⁺ by diaphorase, with the redox dye resazurin being reduced to the highly fluorescent derivative resorufin. The fluorescence of resorufin is stoichiometric with the amount of sorbitol. Retinas (a minimum of one-eighth of a human retina or one whole rat retina) collected in 6% (wt/vol) ice-cold perchloric acid were homogenized by a tissue homogenizer and centrifuged at 20,000g for 15 min. The supernatant was removed and neutralized with 3 mol/l sodium carbonate. The sample was vortexed and centrifuged again as before. An aliquot of the supernatant was added to the assay solution (0.2 mol/l triethanolamine buffer, pH 8.5, 1.2 μmol/l resazurin, 1 mmol/l NAD⁺, and 0.3 units/ml diaphorase with and without 1.0 units/ml sheep sorbitol dehydrogenase). The samples were incubated at room temperature for 1 h, and fluorescence intensity was measured at 580 nm, with the excitation wavelength being 560 nm. Sorbitol concentrations were calculated by comparison with a series of known sorbitol standards. The pellet from the first centrifugation was neutralized with 0.1 N NaOH, briefly sonicated, and rotated overnight at 4°C, and protein was assayed with the BioRad DC protein-assay reagent. Sorbitol levels are expressed as nanomoles per milligram protein.

Data are summarized as means ± SD, except scores and counts, which are summarized with the median and range and presented as box plots because they are noncontinuous variables. ANOVA followed by the Fisher’s protected least significant difference test was used to compare study groups with regard to metabolic data and levels of CD55 and CD59. The nonparametric Kruskal-Wallis followed by multiple comparisons with the Mann-Whitney U test was used to analyze scores and counts. The paired t test was used to analyze sorbitol accumulation under normal- or high-glucose conditions in organ culture.

After both 2.5 and 9 months of diabetes, the diabetic rats showed lower body weight and higher GHb levels than the age-matched controls (Table 1). In the diabetic rats treated with sorbinil, weight and GHb levels were not different from those of the untreated diabetic rats. The amount of insulin required to prevent weight loss was greater after the longer duration of diabetes. Sorbinil consistently prevented cataract formation.

The retinal vessels of rats with 2.5 months of diabetes showed MAC staining, which was almost completely absent from the vessels of control rats (Fig. 1A and B), confirming our previous observations in both rat and human diabetes (19). MAC deposition was much reduced in the retinal vessels of diabetic rats treated with sorbinil (Fig. 1C). In these rats, the masked assessment of frequency and extent of MAC staining yielded a score of 1 (range 0.5–2), which was lower than that in untreated diabetic rats (2 [1–3], P = 0.02), but not different from that in control rats (0.5 [0–1.5]) (Fig. 1D). As previously observed in the human retina (19), in the rat retina the two glycosylphosphatidylinositol-anchored complement inhibitors CD55 and CD59 were also mostly present in vessels. This was documented by comparing vascular versus total retinal enrichment in CD55 and CD59 (Fig. 2A) because we did not find antibodies that stained these proteins satisfactorily in situ in the rat retina. Diabetic rats showed decreased levels of CD55 and CD59 in comparison to controls, and sorbinil prevented the changes (Fig. 2B).

Rats with 9 months of diabetes showed an increased number of apoptotic retinal pericytes, endothelial cells, and total capillary cells when compared with nondiabetic controls, and the increase was prevented by sorbinil (Fig. 3A and B). Apoptosis is expected to decrease the number of pericytes irreversibly because pericytes do not replicate, but it may not affect the number of endothelial cells because the loss is compensated, at least for some time, by increased replication (31). Consistent with this paradigm, there were fewer pericyte nuclei in the diabetic vessels (9 per mm capillary length [range 4–11]) compared with control vessels (12 [9–13]; P = 0.03), while the number of endothelial cell nuclei was similar (12 [4–14] in diabetic retinas and 13 [10–15] in control retinas). In accordance with its prevention of microvascular apoptosis, sorbinil treatment prevented the loss of pericytes (13 [11–14]; P = 0.01 vs. untreated diabetic rats).

Eventually, retinal capillaries become tubes of basement membrane devoid of both pericytes and endothelial cells (Fig. 4A). After 9 months of diabetes, acellular capillaries were already more numerous than in control rats (15 per mm² retina [range 7–22] vs. 6 [2–8]; P = 0.008), and the increase was prevented by sorbinil treatment (5 [3–13]) (Fig. 4B).

The prevention of microvascular cell apoptosis and acellular capillaries by sorbinil generated the question of whether sorbinil could act directly on vascular cells. Aldose reductase immunoreactivity has been reported in retinal pericytes and endothelial cells of the BB rat (14) but not detected in these cell types in other rat studies (13). The methods used for tissue preparation appear to affect aldose reductase immunoreactivity (18). We could not detect signal in the capillary network prepared by trypsin digestion of retinas fixed in 10% formalin for 24–96 h, but we did in capillaries isolated from fresh retinas and fixed in formalin for 15 min (Fig. 5A). Pericytes were positive for aldose reductase, showing rims of cytoplasmic staining around the characteristic round nuclei protruding from the capillary wall. In some areas, the capillary wall stained for both aldose reductase and PECAM-1, an endothelial cell marker (32). Since the dual staining could result from the presence of aldose reductase–positive pericyte processes enveloping PECAM-1–positive endothelial cells, we examined endothelial cells independent of pericytes in dissociated cell preparations (Fig. 5B and C). Cells positive for PECAM-1 were also positive for aldose reductase, indicating that rat retinal endothelial cells contain aldose reductase. No cells in the preparations examined stained for myeloperoxidase (stain performed in the Clinical Hematology Laboratory of the Massachusetts General Hospital), making it unlikely that PECAM-1–positive cells were granulocytes (32).

Human retinal pericytes are known to contain aldose reductase (18), but endothelial cells have been reported not to show immunoreactivity (16,18). The retinal tissue examined in the above studies was fixed in 4% paraformaldehyde for at least 8 h (16,18). Using preparations of dissociated human retinal cells fixed briefly with acetone, we observed (Fig. 6) aldose reductase immunoreactivity in cells showing the granular perinuclear fluorescence of vWf characteristically seen in retinal endothelial cells in situ (24). Hence, both pericytes and endothelial cells of human retinal vessels contain aldose reductase.

Aldose reductase survives for many hours postmortem in several human tissues (33). To establish if aldose reductase in the human retina is enzymatically active, we measured sorbitol in retinas isolated from nondiabetic eye donors and exposed to normal or elevated glucose concentrations in organ culture for 6, 12, 24, and 48 h and compared the response with that exhibited by rat retinas. The peak sorbitol levels were observed after 24 h of incubation in high glucose, at which point the human retinas showed a greater than fourfold increase in sorbitol accumulation (Fig. 7A). The magnitude of the change, as well as the absolute sorbitol levels, were very similar to those observed in rat retinas cultured in parallel. Erythrocytes trapped in retinal vessels were not a major contributor to sorbitol levels. Basal sorbitol levels and the increment in response to high glucose in the human retinas were of the magnitude measured per gram Hb in human erythrocytes (30), whereas the Hb levels in the whole human retina (∼0.1 μg/20 μg retinal lysate) did not exceed 40 μg (Fig. 7B). Thus, cells of the human retina metabolize glucose to sorbitol when exposed to high glucose levels, and the magnitude of sorbitol accumulation is similar to that observed in the rat retina. In both human and rat retinas, there was sorbitol dehydrogenase activity, as indicated by increased levels of fructose after a 24- to 48-h incubation in high glucose (Y.S.P., unpublished observations).

This study supports the concept that, in the rat, activity of aldose reductase is a mechanism for cardinal manifestations of diabetic retinal microangiopathy common to rat and human. We document here that an aldose reductase inhibitor prevented vascular processes culminating in the development of acellular capillaries. Other investigators have shown beneficial effects of different classes of aldose reductase inhibitors on increased vascular permeability (34,35), another clinically important manifestation of diabetic retinal microangiopathy. We reported a preventative effect also on the neuronal apoptosis and glial reactivity (12) found in the diabetic rat retina and occurring in the human retina. These combined observations make the polyol pathway the only mechanism of glucose toxicity currently proven to be responsible for the spectrum of neural, glial, and vascular abnormalities detectable during the development of diabetic retinopathy.

The preventative effect of sorbinil in our studies was not only comprehensive but also noteworthy for three additional reasons. First, the degree of prevention of acellular capillaries was almost complete, while in the same diabetic rat model antioxidant therapy, as tested to date, has shown only partial (36) or absent (37) benefit. Second, prevention was afforded in the presence of severe hyperglycemia, which would be expected to activate several proposed pathways of tissue damage, including excess superoxide production (38), formation of advanced glycation end products (39), and activation of protein kinase C isoforms (40) along with the polyol pathway. Third, prevention of retinopathy occurred despite the fact that aldose reductase, like other members of the aldo-keto reductase superfamily, functions in some circumstances as a detoxification system for lipid peroxidation products (41) that are increased in the retina of diabetic rats (42). Inhibition of the enzyme could have increased toxic products and thus tissue damage, whereas the retinal abnormalities of diabetic rats addressed in our studies were corrected by sorbinil. A likely explanation is that accumulation of lipid peroxidation products in the diabetic rat retina appears to be consequent to polyol pathway activity (42). One should ask if the beneficial effects of sorbinil are solely attributable to inhibition of aldose reductase. The hydantoin class of aldose reductase inhibitors, to which sorbinil belongs, inhibits aldehyde reductase almost as much as aldose reductase (43). However, any contribution of aldehyde reductase to glucose toxicity has yet to be demonstrated and will arguably be low given that the enzyme has a K_m for glucose in the molar range (43). There is no credible evidence that sorbinil and other aldose reductase inhibitors are active in vivo as direct antioxidants (44).

Another novel finding of the present study was the detection of aldose reductase in retinal endothelial cells, both in the rat and in humans. The lack of success of previous attempts may be attributable to the different antibodies used and/or high sensitivity of endothelial (and pericyte) aldose reductase to tissue preparatory steps (fixation in particular). Perhaps the amount of enzyme is smaller in vascular cells than in glial cells or neurons, and any disturbance of the antigenic determinants brings the protein below the threshold of detection by immunohistochemistry. Endothelial cells, which are arbiters of the most consequential lesions of retinopathy, such as capillary permeability and occlusion/obliteration, have been the most reticent among retinal cells to make known how they are affected by diabetes (45). Knowing that they are endowed with aldose reductase offers new ways of probing the effects of diabetes. We can look for a “signature” of polyol pathway activity on the endothelial cells of retinal vessels and eventually reconstruct if endothelial dysfunction and death in diabetic retinopathy is a direct consequence of polyol pathway activation in the endothelium or is attributable to other mechanisms, perhaps extrinsic to the endothelium, such as the death of pericytes or the reactivity of glial cells.

The accumulation of sorbitol by human retinas exposed to high glucose in organ culture is perhaps the first evidence that human retinal aldose reductase is activated under conditions mimicking hyperglycemia. When considering that several factors can influence sorbitol levels, and more may be operative in organ culture, it was noteworthy that the magnitude of changes induced by high glucose in the rat retina was exactly the same as that induced by diabetes (12,42) and that the human retina behaved like the rat retina.

Our data, and the mounting evidence that an allelic variant in the promoter region of the human aldose reductase gene is associated with susceptibility to diabetic retinopathy in both type 1 and type 2 diabetes and in different races (7), renew the rationale for pursuing the role of aldose reductase activation in human diabetic retinopathy. The clinical trial with sorbinil was inconclusive (11), but there was no information that the dose of drug given (20-fold lower than that used in our rat experiments, limited in humans by hypersensitivity reactions) was sufficient to affect polyol pathway activity in situ in the retina. There are precedents for clinical trials failing to prove a plausible hypothesis in diabetes and additional trials proving the hypothesis resoundingly. The University Group Diabetes Program set out to test the “glucose hypothesis” of the cardiovascular complications of diabetes but failed to prove it because the trial was flawed in multiple ways (46). Well-designed and powerful later trials such as the U.K. Prospective Diabetes Study (47) and the Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Complications (48) generated clear evidence in support of the role of hyperglycemia. Also with regard to the “polyol pathway hypothesis” of diabetic retinopathy, the definitive studies will be new trials of aldose reductase inhibitors. It is encouraging that new inhibitors are becoming available (35,42); some with novel structures free from the hydantoin nucleus found in inhibitors such as sorbinil, better tissue penetration and activity than the carboxylic acid inhibitors, and higher selectivity against aldose (versus aldehyde) reductase (49). Thus, there are both new reasons and new probes that encourage testing if the polyol pathway is, after all, a “dream” target in the prevention of diabetic retinopathy and other complications.

The work was supported by the Juvenile Diabetes Research Foundation (JDRF) Center for Diabetic Retinopathy at the Schepens Eye Research Institute, Public Health Service grant EY 09122, and the George and Frances Levin Endowment. Z.D. is the recipient of a JDRF fellowship (3-2003-242).

The authors thank Pfizer Global Research and Development Groton Laboratories for providing sorbinil; Dr. P.J. Oates for many helpful conversations; Dr. T.S. Kern for teaching Z.D. the method to count acellular capillaries; Drs. D.A. Carper, W.G. Couser, B.P. Morgan, and R.L. Sorenson for the gift of antibodies; and the Clinical Hematology Laboratory of the MGH for performing the myeloperoxidase stains.