Aldose Reductase Inhibition Counteracts Oxidative-Nitrosative Stress and Poly(ADP-Ribose) Polymerase Activation in Tissue Sites for Diabetes Complications

The experiments were performed in accordance with regulations specified by the National Institutes of Health Principles of Laboratory Animal Care 1985 Revised Version and the University of Michigan and University of Iowa Protocols for Animal Studies.

Unless otherwise stated, all chemicals were of reagent-grade quality and purchased from Sigma Chemical (St. Louis, MO). Methanol (high-performance liquid chromatography grade), perchloric acid, hydrochloric acid, and sodium hydroxide were obtained from Fisher Scientific (Pittsburgh, PA). Reagents for immunohistochemistry have been purchased from Vector Laboratories (Burlingdale, CA) and Dako Laboratories (Santa Barbara, CA), as specified in the procedures. Bovine aortic endothelial cells (BAECs) and BAEC growth medium were purchased from Cambrex Bio Science Rockland (Rockland, ME) and rat Schwann cells (RCS96) from American Type Culture Collection (Manassas, VA).

Male Wistar rats (Charles River, Wilmington, MA) weighing 250–300 g were fed a standard rat diet (PMI Nutrition Int., Brentwood, MO) and had access to water ad libitum. Diabetes was induced with STZ as previously described (4,17,19). Blood samples for glucose measurements were taken from the tail vein ∼48 h after the STZ injection and the day before the animals were killed. Rats with blood glucose ≥13.8 mmol/l were considered diabetic. The experimental groups comprised control and diabetic rats treated with or without fidarestat (16 mg · kg⁻¹ · day⁻¹, in the diet). The treatments were started immediately after induction of diabetes. The duration of treatment was 6 weeks.

Rats were anesthetized by inactin (65–85 mg/kg body wt i.p.). The onset (before induction of diabetes) motor nerve conduction velocity (MNCV) and sensory nerve conduction velocity (SNCV) measurements were followed by the final (6-week time point) measurements. In all measurements, body temperature was monitored by a rectal probe and maintained at 37°C with a warming pad. Hindlimb skin temperature was also monitored by a thermistor and maintained between 36 and 38°C by radiant heat. After completion of nerve functional measurements, the animals were sedated by CO₂ in a specially designed chamber (26) and immediately killed by cervical dislocation. The left nerve and retina were rapidly dissected, blotted with fine filter paper to remove any accompanying blood, and frozen in liquid nitrogen for subsequent measurements of glucose and sorbitol pathway intermediates. The nerve material was also used for assessment of glutathione (GSH) and ascorbate concentrations as well as fidarestat concentrations (fidarestat-treated control and diabetic groups). The right nerve and retina were fixed in formalin and later used for assessment of nitrotyrosine and poly(ADP-ribose) by immunohistochemistry. In some rats from the control, untreated diabetic, and fidarestat-treated diabetic groups (n = 4–5 per group), sciatic nerves were used for isolation of epineurial arterioles and subsequent assessment of arteriolar superoxide anion radical and nitrotyrosine abundance. Aortas from these rats were also sampled and immediately used for superoxide measurements.

Sciatic nerve epineurial arterioles were isolated as previously described (27).

BAECs were cultured in the BAEC growth medium containing 7 mmol/l glucose, according to the manufacturer’s instructions. Rat Schwann cells 96 (RSC96) were cultured in Dulbecco’s modified Eagle’s medium (5.5 mmol/l glucose) containing 10% fetal bovine serum and a penicillin-streptomycin mixture (Invitrogen, Carlsbad, CA). The final concentrations in the culture medium are 100 μg/ml streptomycin and 100 units/ml penicillin. We used passages 5–12 for experiments with both cell types. Sciatic MNCV and hindlimb SNCV measurements were performed as described (4).

Glucose, sorbitol, and fructose concentrations in sciatic nerve and retina were assessed spectrofluorometrically by enzymatic procedures with hexokinase/glucose 6-phosphate dehydrogenase, sorbitol dehydrogenase, and fructose dehydrogenase as described (26). GSH and ascorbate concentrations in the sciatic nerve have been measured spectrofluorometrically with O-phthaldialdehyde and O-phenylenediamine, respectively (4).

All immunohistochemical samples were coded and examined by a single investigator in a blinded fashion. Microphotographs of stained sciatic nerves and retinas were taken with a Zeiss Axiolab microscope equipped with a Fuji HC-300C digital camera.

Sciatic nerves and eyes were fixed in 4% paraformaldehyde in PBS, and 5-μm sections were prepared from paraffin-embedded tissues. Endogenous peroxidase was quenched with 0.3% H₂O₂ in 60% methanol for 15 min. The sections were incubated overnight with 1:1,000–1:2,000 dilution of primary anti-nitrotyrosine antibody (Upstate Biotechnology, Lake Placid, NY). In control measurements, tissues were incubated with the primary antibody in the presence of 10 mmol/l nitrotyrosine. Specific labeling was detected with a biotin-conjugated goat anti-rabbit IgG and avidin-biotin peroxidase complex, both of which were supplied in the Vector Elite kit (Vector Laboratories). Color was developed using the Ni-diaminobenzidine substrate kit (Vector Laboratories). Sections were counterstained with hematoxylin-eosin, dehydrated, and mounted in Permount. The photomicrographs shown are representative sections (n = 5–6) for each experimental group. A similar staining procedure has been used for epineurial vessel sections.

Paraffin sections (5 μm) were loaded onto polysine-coated slides (Fisher, Atlanta, GA), deparaffinized, and rehydrated. Optimal staining was achieved with an antigen retrieval method that was performed in 10 mmol/l citric acid for 15 min. Endogenous peroxidase was quenched with 0.3% H₂O₂ in 60% methanol for 15 min. Sections were blocked with 2% normal goat serum at room temperature for 1–2 h and incubated overnight with 1:250–1:500 dilution of primary anti-poly(ADP)-ribose antibody (generous gift from Tulip Biolabs). Specific labeling was detected with a biotin-conjugated goat anti-chicken IgG and avidin-biotin peroxidase complex (Vector Laboratories). The enzymatic reaction product was enhanced with nickel cobalt to give a black precipitate, and the sections were counterstained with hematoxylin-eosin, dehydrated, and mounted in Permount. Positive controls included formalin-fixed, paraffin-embedded tissues from lipolysaccharide-treated rats. Negative controls included elimination of the primary antibody.

Superoxide anion radical (O₂⁻) abundance in epineurial vessels was assessed by the hydroethidine method, as we have described in detail (27). Vessels from control rats and untreated and treated diabetic rats were processed and imaged in parallel. Laser settings were identical for acquisition of all images from control and diabetic specimens.

Superoxide anion radical (O₂⁻) abundance in aorta was measured by lucigenin-enhanced chemiluminescence, as previously described (27). Vessel segments were incubated in 0.5 ml PBS containing lucigenin (5 μmol/l); afterward, relative light units (RLUs) were measured using a Zylux FB12 luminometer. For these studies, chemiluminescence was measured for 5 min. Background activity was determined and subtracted, and RLUs were normalized to surface area.

The BAECs were cultured (a six-well plate per condition) in the BAEC growth medium containing 7 mmol/l glucose, 30 mmol/l glucose, or 30 mmol/l glucose plus 1 μmol/l fidarestat (added for 4 h at the end of experiment). The total duration of the experiment was 4 days. Ten microliters of 10 μmol/l CM-H₂DCFDA, the dichlorofluorescein derivative with the best retention properties among all the studied analogs, was added for 30 min at the end of the experiment. The cells were then washed and trypsinized, and CM-H₂DCFDA fluorescence, an index of reactive oxygen species generation, was measured by flow cytometry (λ excitation, 480 nm; λ emission, 520 nm). After flow cytometry, the BAECs were counted and CM-H₂DCFDA fluorescence expressed per 10⁶ cells.

For detection of poly(ADP-ribose) formed in Schwann cells, ∼9,000 RSC96 in their growth media were seeded in each well of a 96-well enzyme-linked immunosorbent assay plate. After 24 h, the media were discarded and replaced by the new media containing 5.5 mmol/l glucose, 30 mmol/l glucose, or 30 mmol/l glucose plus 1 μmol/l fidarestat, for another 8 h, before CELISA (cell enzyme-linked immunosorbent assay) using the Poly(ADP-ribose) Polymerase Activity Assay Kit (Trevigen, Gaithersburg, MD). For detection of poly(ADP-ribose) formed in cell-free system, wells were coated with 1 mg/ml histone (50 μl/well) at 4°C overnight. Then, they were washed four times with PBS and blocked by adding 50 μl Strep-Diluent (supplied with the Trevigen kit). After incubation (1 h at room temperature), plates were washed four times with PBS. Appropriate dilutions of fidarestat were combined with a 2× PARP cocktail (1.95 mmol/l NAD⁺, 50 μmol/l biotinylated NAD⁺ in 50 mmol/l Tris, pH 8.0, and 25 mmol/l MgCl₂) and high-specific activity PARP enzyme (both supplied with the Trevigen kit) in a 50-μl volume. The reaction was allowed to proceed for 30 min at room temperature before CELISA, which was performed according to the manufacturer’s instructions. The results have been expressed as a percentage of the activity observed with no fidarestat present.

The homogenate was prepared by disrupting frozen segments of rat sciatic nerve (∼10–15 mg) in a mortar with distilled water. Internal standard solution of 8-fluoro-fidarestat was added to nerve homogenates. For calibration standards, a standard solution of fidarestat and internal standard was added to blank nerve homogenates. The homogenates were extracted four times with ethyl acetate, and the organic layers were evaporated to dryness. The residues were dissolved in the liquid chromatography–tandem mass spectrometry mobile phase, i.e., 0.2% AcOH/CH₃CN (7:3). The solutions were filtered and fidarestat concentrations quantified using a liquid chromatography–tandem mass spectrometer (TSQ7000; Thermo Electron, Finnigan, MA).

The results are expressed as means ± SE. Data were subjected to equality of variance F test and then to log transformation, if necessary, before one-way ANOVA. Where overall significance (P < 0.05) was attained, individual between-group comparisons were made using the Student-Newman-Keuls multiple-range test. Significance was defined at P < 0.05. When between-group variance differences could not be normalized by log transformation (datasets for body weights, plasma glucose, and some metabolic parameters), the data were analyzed by nonparametric Kruskal-Wallis one-way ANOVA followed by the Bonferroni/Dunn test for multiple comparisons.

The final body weights were comparably lower in untreated and fidarestat-treated diabetic rats than in the control group (Table 1). The final blood glucose concentrations were similarly elevated in untreated and fidarestat-treated diabetic rats compared with control rats.

Glucose, sorbitol, and fructose concentrations in sciatic nerve and retina were increased in diabetic rats compared with controls (Table 2). Fidarestat treatment did not affect glucose concentrations but essentially normalized sorbitol and fructose concentrations in sciatic nerve and retina of diabetic rats.

The onset MNCVs and SNCVs (both in m/s) were 54.7 ± 0.6 and 36.3 ± 0.3 in control rats, 55.1 ± 0.7 and 36.6 ± 0.4 in control rats treated with fidarestat, 54.6 ± 0.7 and 36.2 ± 0.4 in diabetic rats, and 54.3 ± 0.7 and 36.4 ± 0.5 in diabetic rats treated with fidarestat. The final MNCVs and SNCVs were reduced in diabetic rats compared with controls. Fidarestat treatment prevented both MNCV and SNCV slowing in diabetic rats without affecting either variable in control rats (Fig. 1). Sciatic nerve GSH and ascorbate concentrations were decreased in diabetic rats compared with controls (Fig. 2). Fidarestat treatment prevented nerve GSH and ascorbate depletion in diabetic rats without affecting those variables in control rats.

The sciatic nerve fidarestat concentration was 4.41 ± 0.23 nmol/g wet wt in fidarestat-treated control rats and 4.59 ± 0.29 nmol/g wet wt in fidarestat-treated diabetic rats.

Nitrotyrosine immunoreactivities were increased in sciatic nerve and retina of diabetic rats compared with controls, and this increase was markedly reduced by fidarestat treatment in both tissues (Fig. 3A). In a similar fashion, the diabetes-associated increase in poly(ADP-ribose) immunoreactivity in sciatic nerve and retina was less manifest in diabetic rats treated with fidarestat than in untreated diabetic rats (Fig. 3B).

Superoxide and nitrotyrosine immunoreactivities were increased in epineurial vessels of diabetic rats compared with controls (Figs. 4A and B). The diabetes-associated increase in both immunoreactivities was markedly prevented by fidarestat treatment.

Superoxide abundance in aorta was significantly increased in diabetic compared with control rats (2.95 ± 0.21 and 1.64 ± 0.20 RLU, respectively, P < 0.01), and this increase was prevented by fidarestat treatment (1.52 ± 0.16 RLU, P < 0.01 vs. untreated diabetic group).

CM-H₂DCFDA fluorescence was increased in BAECs cultured in 30 mmol/l glucose compared with those cultured in 7 mmol/l glucose (Fig. 5A). This increase was completely corrected by fidarestat. Poly(ADP-ribose) abundance was increased in RSC96 cultured in 30 mmol/l glucose compared with those cultured in 5.5 mmol/l glucose (Fig. 5B), and this increase was prevented by fidarestat. Fidarestat, in the concentration range of 10 nmol/l to 100 μmol/l, did not affect PARP activity in the cell-free system containing PARP and NAD⁺ (Table 3).

Numerous findings implicate increased aldose reductase activity in the pathogenesis of diabetic neuropathy (3–5,8,10,22), retinopathy (3,5,19,24,25,28), nephropathy (3), and macrovascular disease (3). The role for aldose reductase in the pathogenesis of diabetes complications is supported by, at least, four lines of evidence: 1) similarity of physiological, metabolic, and morphological abnormalities in animal models of diabetes and galactose feeding (29,30); 2) prevention, reversal, or slowing of diabetes complications by structurally diverse ARIs (3–5,19,22,28,30); 3) findings in cell culture models indicating that aldose reductase overexpression causes and aldose reductase inhibition corrects hyperglycemia-induced metabolic and gene expression abnormalities (31,32); and 4) aldose reductase gene polymorphism data obtained in human subjects with diabetes (3). The role for aldose reductase in the pathogenesis of diabetic neuropathy is also supported by 1) the presence of more severe diabetic and diabetes-like neuropathy in aldose reductase–overexpressing mice compared with the wild-type mice with normal aldose reductase content (33); 2) clinical trials of the ARIs zenarestat (34) and fidarestat (35), indicating that robust aldose reductase inhibition in patients with diabetes improves peripheral nerve function and morphology; and 3) identification of a high–aldose reductase protein level as an independent risk factor for diabetic neuropathy in patients with both type 1 and type 2 diabetes (3). The mechanism(s) of aldose reductase involvement in the pathogenesis of diabetes complications is not completely understood. Early studies revealed that increased aldose reductase activity leads to sorbitol accumulation and resulting osmotic stress (16,36). Later, increased aldose reductase activity has also been found to result in a variety of biochemical abnormalities, including myo-inositol depletion and downregulation of Na/KATP-ase activity (32,37), NAD⁺/NADH and NADP⁺/NADPH redox imbalances (4,17), changes in fatty acid metabolism (38), impaired neurotrophic support (39), and upregulation of vascular endothelial growth factor (19). Some of these disturbances (myo-inositol depletion) are obviously related to and some [NAD(P)⁺/NAD(P)H redox changes] are independent of intracellular osmotic stress. Here, we demonstrate that increased aldose reductase activity is a major contributor to oxidative-nitrosative stress and PARP activation in diabetic peripheral nerve and retina.

Our animal studies have documented increased superoxide production in epineurial vessels of diabetic rats, and in vitro experiments with BAECs confirmed the endothelial origin of this phenomenon. Increased superoxide production in vasa nervorum correlated with that in large vessels (aorta). Aldose reductase inhibition with fidarestat clearly prevented diabetes-associated superoxide formation in both aorta and epineurial vessels. These findings are consistent with several other reports indicating that structurally diverse ARIs, fidarestat (19) and zopolrestat (20,40), prevent hyperglycemia-induced superoxide formation (assessed by several different methods [19,20,40]) in aorta and retinal endothelial cells. We have also documented the presence of peroxynitrite-induced injury in peripheral nerve, epineurial vessels, and retina in diabetic rats. Peroxynitrite (ONOO⁻) is a potent oxidant that causes 1) initiation of lipid peroxidation; 2) DNA breakage and base modification; 3) protein (tyrosine amino acid) nitration and nitrosylation; 4) alterations in cell signaling; and, 5) in cases of severe damage, induction of cell necrosis and apoptosis (41). Nitrotyrosine immunoreactivities in sciatic nerve, epineurial vessels, and retina were reduced by fidarestat treatment. The latter is quite understandable, considering that peroxynitrite is a product of superoxide anion radical reaction with nitric oxide and that aldose reductase inhibition decreases superoxide abundance.

Growing evidence indicates that increased aldose reductase activity leads to diabetes-associated oxidative-nitrosative stress via several mechanisms. First, aldose reductase is responsible for downregulation of antioxidative defense provided by both nonenzymatic antioxidants, i.e., GSH, ascorbate (4,16,17 and current study), and taurine (31), and antioxidative defense enzymes, i.e., superoxide dismutase, catalase, glutathione peroxidase, glutathione reductase, and glutathione transferase (10,19). The mechanisms underlying this phenomenon are not clear. Both osmotic stress (10,16,26) and NADPH deficiency (33,42) have been implicated in diabetes-induced GSH depletion in lens and peripheral nerve. However, on the one hand, some findings in aldose reductase–transgenic mice are not consistent with the osmotic origin of oxidative stress (42). On the other hand, NADPH deficiency and related increase in oxidized glutathione (GSSG) concentration (a half-molar equivalent of GSH depletion) have never been experimentally documented in tissue sites for diabetes complications. In reality, GSSG concentrations in peripheral nerve, retina, and kidney are very low (<5% of GSH concentrations [10,17,19]) and diabetes-associated increase is either minimal or absent (10,17). Osmotic stress is clearly implicated in the depletion of another important nonenzymatic antioxidant, taurine (31).

In addition to disruption of antioxidative defense, increased aldose reductase activity contributes to oxidative-nitrosative stress indirectly via other pathways. Aldose reductase is involved in the generation of fructose, a glycation agent that is 10 times more potent than glucose. Increased aldose reductase activity results in formation of other precursors of advanced glycation end products (AGEs), i.e., methylglyoxal and 3-deoxyglucosone (43,44), and the AGEs pentosidine and carboxymethyllysine per se (45,46). Several reports indicate that aldose reductase inhibition counteracts formation of AGEs (45,46), which are known to generate oxidative stress via interaction with their receptors (47). Thus, increased aldose reductase activity contributes to oxidative stress by promoting nonenzymatic glycation. So far, several groups have reported that increased aldose reductase activity leads to activation of PKC in glomerular mesangial cells (48), vasculature (49), and smooth muscle cells (50). The latter probably occurs due to a aldose reductase–dependent decrease in free cytosolic NAD⁺/NADH ratio and associated diversion of the glycolytic flux toward the increased formation of α-glycerophosphate and -diacylglycerol, an activator of PKC. PKC is essentially required for phosphorylation (activation) of NAD(P)H oxidase, the superoxide-generating enzyme (51), which is particularly important in the vasculature (8,9,51). Therefore, increased aldose reductase activity can contribute to increased superoxide generation via activation of PKC.

The present study also provides evidence that increased aldose reductase activity is responsible for PARP activation in the diabetic nerve and retina as well as in high-glucose–exposed Schwann cells. PARP activation, a recently discovered fundamental mechanism in the pathogenesis of diabetic endothelial (13) and myocardial (52) dysfunction, peripheral neuropathy (14), and retinopathy (53), is completely or partially prevented in the two tissues of diabetic rats treated with the ARI fidarestat. Aldose reductase colocalizes with PARP in both tissue targets for diabetes complications. In the retina, aldose reductase and PARP are localized primarily in the neural retina (28,53) (Fig. 2B). In the sciatic nerve, aldose reductase is primarily localized in Schwann cells (33), and PARP, as we have recently demonstrated using the Schwann cell marker S-100 (14), is primarily localized in the Schwann cell nuclei. PARP is also known to be present in endothelial cells (13,54,55), including peripheral nerve endothelial cells (14), and evidence for endothelial localization of aldose reductase is emerging (28). However, the vasculature constitutes only ∼1% of the total rat retina and <5% of the sciatic nerve. Therefore, it is reasonable to assume that 1) the vast majority of poly(ADP-ribosyl)ated proteins is contained in the neural elements of both tissues and 2) the effect of aldose reductase inhibition on poly(ADP-ribose) accumulation detected by immunohistochemistry is primarily localized in neuronal and glial retinal cells and in peripheral nerve Schwann cells.

It has been considered for many years that hydantoin ARIs may have weak antioxidant properties (56). Using bovine retinal endothelial cells, we previously found that fidarestat does not affect oxidative stress caused by factors other than hyperglycemia, i.e., three different pro-oxidants (19). The weak antioxidant capacity, if any, cannot explain the effects of fidarestat on diabetic neuropathy and retinopathy because conventional antioxidants, and even such potent compounds as DL-α-lipoic acid, require administration at much higher doses to produce effects on two complications’ end points and oxidative stress in diabetic retina and peripheral nerve (57,58). Here, we show that fidarestat does not directly inhibit PARP in the cell-free system containing PARP and NAD⁺. Note that we have tested the PARP-inhibiting capacity of fidarestat with very high concentrations of the compound, i.e., up to 100 μmol/l. This concentration exceeds the tissue (peripheral nerve) concentrations in diabetic rats by >20-fold and the concentration used for Schwann cells by 100-fold.

Growing evidence suggests that aldose reductase inhibition and PARP inhibition similarly affect the key biochemical abnormalities developing in diabetic and hyperglycemic conditions. Both aldose reductase inhibition and PARP inhibition normalize peripheral nerve mitochondrial and cytosolic NAD⁺/NADH ratios and energy deficiency (4,59), as well as retinal VEGF overexpression in STZ-induced diabetic rats (19,53). As we have discussed above, aldose reductase inhibition is known to prevent diabetes-induced diversion of the glycolytic flux from glyceraldehyde 3-phosphate dehydrogenase toward the formation of α-glycerophosphate (10), the PKC activator diacylglycerol (48), and methylglyoxal (43). All these changes are prevented by structurally diverse ARIs (10,43,48). Increased formation of methylglyoxal and PKC activation has also been recently found to occur under conditions of high-glucose–induced PARP activation and to be preventable by PARP inhibition (55), further supporting the link between increased aldose reductase activity and PARP activation in the pathogenesis of diabetes complications. We have not found any suppression of sorbitol pathway hyperactivity in either diabetic retina (53) or peripheral nerve (59) by three structurally diverse PARP inhibitors, which is consistent with the downstream localization of the oxidative stress–triggered PARP activation consequent to increased aldose reductase activity, in the pathogenesis of diabetic neuropathy and retinopathy.

Our present findings are consistent with other studies demonstrating prevention of MAPK activation with aldose reductase inhibition by the ARIs AL-1576 (60) and fidarestat (23). Increased aldose reductase activity is the most likely to lead to MAPK activation via oxidative stress because it is well known that reactive oxygen and nitrogen species cause activation of all three major subfamilies of MAPKs (12). Both PARP activation and MAPK activation are involved in transcriptional regulation of gene expression via the transcription factors nuclear factor κB, activator protein-1, p53, and others (61,62). Activation of these transcription factors leads to upregulation of inducible nitric oxide synthase, cyclooxygenase-2, endothelin-1, cell adhesion molecules, and inflammatory genes (61,63). Growing evidence indicates that all of these genes are involved in the pathogenesis of diabetes complications (15,64,65). The demonstration of a major contribution of aldose reductase to oxidative-nitrosative stress and PARP activation in tissue sites for diabetes complications allows us to predict that, in the near future, the link between increased aldose reductase activity and altered transcriptional regulation and gene expression will be established. It is important to realize that any product of genes controlled via PARP- and MAPK-dependent transcription factors, regardless of how unrelated it appears to be (from a biochemical viewpoint) to the sorbitol pathway, will be affected by a diabetes-associated increase in aldose reductase activity and amenable to control by aldose reductase inhibition.

In conclusion, our study generated evidence for the major role of increased aldose reductase activity in diabetes-associated oxidative-nitrosative stress and PARP activation. The data provide a rationale for studying the role for aldose reductase in diabetes-associated altered cell signaling, transcriptional regulation, and gene expression. The results reveal new beneficial properties of fidarestat, thus further justifying the ongoing clinical trials of this specific, potent, and low-toxic ARI.

The study was supported by an American Diabetes Association research grant (to I.G.O.), National Institutes of Health Grant DK59809-01 (to I.G.O.), Juvenile Diabetes Research Foundation Center for the Study of Complications of Diabetes Grant 4-200-421 (to I.G.O. and M.J.S.), National Institutes of Health Grant HL/DK 71215-01 (to C.S.), and National Institutes of Health Grant DK 058005-04 (to M.A.Y.).

The authors are grateful to Dr. Joseph Eichberg for valuable recommendations regarding the RSC96 cell culture experiments.