RAGE Control of Diabetic Nephropathy in a Mouse Model: Effects of RAGE Gene Disruption and Administration of Low–Molecular Weight Heparin

To perform gene targeting in embryonic stem cells, we constructed a targeting vector containing two loxP sites flanking the neo cassette in intron 2 and another loxP that was inserted into 0.6 kilobases (kb) 5′ upstream of exon 1 that had the translation initiation site (Fig. 1A). After gene targeting in the E14-1 embryonic stem cells, we identified six targeted clones by PCR and Southern blotting with probe 1, of which two were used for further experiments. The two clones containing all three loxP sites in the locus gave rise to germ line chimeras by the aggregation method (10). The resultant male chimeras were mated with female Cre-transgenic mice (CD-1 background), which transiently express Cre recombinase in eggs (11). Some of the newborn mice were found to carry the deleted allele that lacks both RAGE exons 1 and 2 and neo cassette. Mutant mice were backcrossed to CD-1 for more than four generations. For PCR genotyping, three primers were used: 5′-CCAGAGTGACAACAGAGCAGAC-3′ (primer 1), 5′-GGTCAGAACATCACAGCCCGGA-3′ (primer 2), and 5′-CCTCGCCTGTTAGTTGCCCGAC-3′ (primer 3) (nucleotides 73915-73936, 74523-74544, and 74881-74902 in GenBank accession no. AF030001, respectively). To induce diabetes, the mutant RAGE (+/−) mice were crossbred with other transgenic mice (CD-1) carrying human cDNA for inducible nitric oxide synthase (iNOS) under the control of an insulin promoter (iNOSTg) (8,12). The resultant male iNOSTg (+) RAGE (+/−) mice were crossbred with female RAGE (+/−) mice. The resultant six groups of male littermates were used for analysis after PCR verification. All animals were fed a standard mouse diet (346 kcal/100 g [protein 30.4, fat 8.6, and carbohydrate 44.0%], Labo H Standard; NIHON NOSAN, Kanagawa, Japan). The levels of blood glucose and HbA_1c (A1C) were measured from tail vein blood using a Dexter Z sensor and DCA2000 analyzer (Bayer-Medical, Tokyo, Japan) (8), respectively. The procedures were approved by the institutional animal care and use committee guideline at Kanazawa University.

Total RNA was isolated from various tissues using a high pure RNA isolation kit (Roche Diagnostics, Basel, Switzerland) and reverse transcribed as described (8,13). The primer sequences for mouse RAGE mRNA detection were 5′-CCTGGGTGCTGGTTCTTGCTCT-3′ and 5′-GATCTGGGTGCTCTTACGGTCC-3′ (nucleotides 31-52 and 12091-230 in GenBank accession no. L33412); those for mouse glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA detection, human vascular endothelial growth factor (VEGF) mRNA detection, and human β-actin mRNA detection were the same as described (8,14,15). The amounts of total RNA templates (100 ng) and the numbers for amplification cycles (30 cycles for mouse RAGE, GAPDH, and human VEGF; 28 cycles for human β-actin) were chosen in semiquantitative ranges. An aliquot of each RT-PCR product was electrophoresed on 2% agarose gel containing ethidium bromide. Western blot analysis was carried out as previously described (8,16). Briefly, proteins in the tissue lysates were boiled and resolved by SDS-PAGE (12.5%) and then transferred onto a polyvinylidene fluoride membrane (Millipore). The membranes were incubated with a polyclonal anti-RAGE antibody, and immunoreacted bands were visualized with an enhanced chemiluminescence detection system (Amarsham Pharmacia Biotech).

Serum N^ε-carboxymethyl-lysine (CML) and non-CML AGEs were differentially determined using a competitive enzyme-linked immunosorbent assay (ELISA) with an anti-CML antibody and an anti–non-CML AGE antibody as previously described (8,17). The non-CML AGE antibody used in this study did not cross-react with several structurally identified AGE-modified BSA, including pyrraline BSA, pentosidine BSA, argpyrimidine BSA, 3-deoxyglucosone imidazolone BSA, carboxymethyllysine BSA, carboxyethyllysine BSA, glyoxal-lysine dimmer, or methyglyoxal-lysine dimmer (18); undefined AGE structures may be recognized by the non-CML AGE antibody. CML or non-CML AGE (1 unit/ml) corresponded to a protein concentration of 1 μg/ml CML-BSA or non-CML AGE-BSA, respectively.

Urinary or serum creatinine was measured by the Jaffe reaction (8,19). The level of urinary albumin was measured by ELISA with sheep anti-mouse albumin as described (8,20). The ratio of urinary albumin to creatinine was then calculated.

Kidneys were processed for light microscopic examination, and the severity of renal sclerosis was scored by plural analysis on an arbitrary 0- to 4-point scale (8,21,22). The mean glomerular tuft area was determined as described previously (8). To quantify glomerulosclerosis, periodic acid-methenamine silver (PAM) stain-positive area in the mesangium was determined using an image analyzer with a microscope (IPAP; Sumitomo Chemical, Osaka, Japan) (23). Immunofluorescence analysis was carried out as previously described using an anti-CML antibody and an anti–non-CML AGE antibody (8,24).

We used dalteparin sodium (Fragmin; Pfizer) as a LMWH, which was produced from porcine intestine–derived heparin by the deaminative hydrolysis with nitrous acid. Its average molecular weight is 5,000. Direct binding of LMWH to RAGE was assayed by surface plasmon resonance (SPR) assay (Biacore 2000; Biacore, Uppsara, Sweden). Purified secretory isoform of RAGE proteins (16), which have a ligand-binding site, were immobilized on a CM5 research-grade sensor chip according to standard amine coupling procedures (15,25) with an amine coupling kit (Biacore) to a density of ∼5,500 response units (RU); this density corresponded to 110 fmol/mm². On the control surface, rabbit IgG was immobilized at the same density. The sample and flow buffer used was 10 mmol/l HEPES (pH 7.4), 0.15 mol/l NaCl, 3 mmol/l EDTA, and 0.005% surfactant P-20, and binding was measured at 25°C at a flow rate of 20 μl/min. The sensor chips were regenerated by washing with 10 mmol/l NaOH and 0.5% SDS. The mean equilibrium dissociation constant (K_d) was measured using global fitting of monoexponential rate equations derived from the simple 1:1 Langmuir binding model. Competitive inhibition assay by LMWH in AGE-RAGE interaction was performed using an AGE-BSA–coated 96-well plate method as previously described (26).

C6 rat glioma cells were used for this assay according to a previously described procedure (25,27). Briefly, cells were transfected with a plasmid encoding luciferase cDNA under an enhancer element containing five nuclear factor-κB (NFκB)-binding sites (pNFκB-Luc; Stratagene) and a plasmid containing human full-length RAGE cDNA. Single stably transfected clones were selected, and the expression of RAGE was verified by Western blotting. A typical clone was used for subsequent analyses. After a 24-h preincubation in Dulbecco’s modified Eagle’s medium, supplemented with 0.1% fetal bovine serum, the cells were stimulated by glyceraldehyde-derived AGE-BSA (16) for 4 h. Endotoxin was tested with Limulus HS-test Wako (Wako Pure Chemical Industries, Osaka, Japan); endotoxin was not detected in the AGE preparations. The luciferase activity was determined using Luciferase Assay System (Promega) and measured in a luminometer (Fluoroscan Ascent FL; Labsystems). RAGE expression was silenced by the small interfering RNA (siRNA)-expression system using the pSilencer 3.0-H1 vector (Ambion, Austin, TX). A dominat-negative RAGE lacking cytoplasmic domain was overexpressed according to a previous report by Huttunen et al. (27).

Vascular cell adhesion molecule (VCAM)-1 assay was performed by ELISA as described (28). Briefly, human umbilical vein endothelial cells (HUVECs) were seeded on a 96-well plate and were stimulated with glyceraldehydes-derived AGE-BSA or nonglycated BSA 10 μg/ml for 16 h after a 1-h pretreatment with LMWH 0.1 or 1.0 IU/ml. After the stimulation, the cells were fixed with 1% paraformaldehyde for 15 min and then blocked overnight at 4°C with 2% BSA. After a 2-h incubation with monoclonal antibodies to VCAM-1 (R&D Systems), the cells were incubated for 2 h with 1/1,000 diluted anti-mouse IgG conjugated to alkaline phosphatase (Zymed, San Francisco, CA). Quantification was performed by determination of the colorimetric conversion at optical density 405 nm of substrate p-nitrophenyl phosphate (PIERCE, Rockford, IL). The degree of specific antibody binding was calculated by subtracting the mean negative value without primary anti–VCAM-1 antibody. For the VEGF gene assay, RAGE overexpressing ECV304 cells (the HUVEC-derived cell line) were used. These cells were stimulated with 50 μg/ml of glyceraldehyed-derived AGE-BSA for 4 h with or without LMWH pretreatment. Total RNA was isolated using a total RNA isolation kit (Qiagen), and VEGF gene expression was analyzed by quantitative RT-PCR as previously described (15).

For the prevention study, male iNOSTg mice were used and classified into four groups: one group receivedLMWH via osmotic pump (0.25 IU/h delivery rate; Alzet), and the other groups received subcutaneous (s.c.) injection of 40 or 80 IU of LMWH and vehicle (PBS) once a day. Wild-type CD-1 littermate mice were used as the nondiabetic control. These mice were treated from 1 to 4 months of age. For the study of therapeutic effects of LMWH, we used 4-month-old male iNOSTg mice, which were evaluated by renal tissue biopsy to have overt diabetic nephropathy before the LMWH treatment. Plasma LMWH concentrations and activated partial thromboplastin time were measured as described (29,30).

Formalin-fixed, paraffin-embedded kidney blocks were cut into 4-μm-thick sections and processed for immunohistochemistry as described (31). The primary antibodies used are a rabbit polyclonal VEGF antibody (dilution 1:50; Santa Cruz Biotechnology, Santa Cruz, CA), a rabbit transforming growth factor-β1 polyclonal antibody (dilution 1:50; Santa Cruz Biotechnology), a goat connective tissue growth factor (CTGF) polyclonal antibody (dilution 1:50; Santa Cruz Biotechnology), and a rabbit S100 polyclonal antibody (dilution 1:400; Dako). The peroxidase-labeled polymer for rabbit polyclonal antibody (EnVision; Dako, Carpineria, CA) or goat polyclonal antibody (Histofine Simple Stain; NICHIREI, Tokyo, Japan) was applied as a secondary antibody. Then, color was developed with 3-3′-diaminobenzidine (Sigma). The sections were counterstained with Meyer’s hematoxylin and cover slipped for microscopy observation. Brown areas were judged as positive based on the manufacturing information from Envision System.

All values were expressed as the means ± SE. Comparisons among the groups were analyzed by Student’s t test or ANOVA combined with a multiple comparison test (Scheffe’s type). These analyses were carried out with the use of Stat View V software (SAS Institute). P < 0.05 was considered significant.

To generate RAGE-null mice, a targeting vector was constructed (Fig. 1A) and used for isolating embryonic stem cell clones that acquired the targeting allele by homologous recombination (Fig. 1B). We next created chimeric mice composed predominantly of the targeted embryonic stem cells. The DNA fragment flanked by loxP sequences was removed by a transient expression of Cre recombinase in fertilized eggs. Newborn mice were found to carry the mutated allele that lacked RAGE exons 1 and 2 and neo cassette (Fig. 1C). Crossbreeding between the RAGE (+/−) heterozygous mice resulted in the production of RAGE (−/−) homozygous mice (Fig. 1C and D). The RAGE (−/−) mice were born in accordance with Mendelian laws and grew apparently normal without gross phenotypic or microscopic abnormalities. RT-PCR and Western blot analyses, respectively, revealed the absence of RAGE mRNA and protein production in the RAGE (−/−) mice (Fig. 1E and F), indicating that the gene disruption resulted in a null mutation of RAGE. The same genetic mean was used to induce diabetes, as was the case with the previously reported RAGE transgenic mice: iNOS transgenic mice, which consistently develop hypoinsulinemic diabetes as early as 1 week after birth due to the NO-mediated selective destruction of insulin-producing pancreatic β-cells (8,12). These mice represented a stable diabetic state with enhanced AGE formation and accumulation. After being backcrossed for at least four generations to CD-1, which is the background ensuring a stable induction of diabetes, RAGE (−/−) mice were crossbred with iNOSTg mice (CD-1 background). The resultant male heterozygous iNOSTg/RAGE (+/−) were then mated with female RAGE (+/−) to yield six groups of littermates with the CD-1 background. Three groups carrying the iNOS transgene developed diabetes and were designated as DM⁺RAGE (+/+), DM⁺RAGE (+/−), and DM⁺RAGE (−/−). The other three devoid of the iNOS transgene never developed diabetes and were designated as DM^-RAGE (+/+), DM^-RAGE (+/−), and DM^-RAGE (−/−). Blood analysis confirmed sustained hyperglycemia and high A1C levels in the diabetic mice, but there were no significant differences among the three groups (Table 1). Body weight also did not significantly vary among the three diabetic groups at all time points tested during the 8-month observation period (Table 1). To determine the levels of circulating AGE, serum was assayed by competitive ELISA for the CML derivative of AGE, mainly formed by peroxidation, and for non-CML AGE (8,17). Serum CML and non-CML AGE levels significantly increased in the DM⁺RAGE (+/+) mice in comparison with those in the DM^-RAGE (+/+) mice at 8 months of age, but there were no significant differences among the three diabetic groups (Table 1). On the other hand, serum non-CML AGE was the highest in DM^-RAGE (−/−) mice among the nondiabetic groups, suggesting that AGE might be accumulating as the animals age when RAGE is totally depleted (Table 1).

Nephromegaly, expressed as an increase in weight ratio of the kidney against the body, was significantly attenuated by RAGE depletion at both 4 (Fig. 2A and B) and 8 months of age (Fig. 3B). Urinary albumin-to-creatinine ratio was significantly decreased in the DM⁺RAGE (−/−) mice when compared with the DM⁺RAGE (+/+) mice at 4 months of age (Fig. 2C). The microscopic lesions noted in the DM⁺RAGE (+/+) mice consisted of glomerular cell proliferation, mesangial expansion, and glomerulosclerosis (Figs. 2D and 3A). Quantitatively, glomerular cell number was significantly smaller in the DM⁺RAGE (−/−) mice than the DM⁺RAGE (+/+) mice at 4 and 8 months of age (Figs. 2E and 3C). Quantitative evaluation of glomerulosclerosis by calculating the sclerosis index and PAM stain–positive area per glomerular tuft area revealed that progression of the sclerosis was significantly attenuated in the DM⁺RAGE (−/−) mice compared with the DM⁺RAGE (+/+) mice (Figs. 2G and H and 3E and F); it should be noted that the values obtained with the DM⁺RAGE (+/−) mice were consistently intermediate to those with the DM⁺RAGE (+/+) and DM⁺RAGE (−/−) mice. At 8 months, the increase in serum creatinine level became evident in the DM⁺RAGE (+/+) mice; this was also significantly suppressed in the DM⁺RAGE (+/−) and DM⁺RAGE (−/−) mice (Fig. 3G). In addition, we examined AGE accumulation at 4 months by immunostaining. Non-CML or CML AGE was more densely deposited in the renal glomeruli of the DM⁺RAGE (+/+) mice than in those of the RAGE-null mutants (Fig. 2I). Expression of S100 protein, one of RAGE ligands, was also the highest in the glomeruli of DM⁺RAGE (+/+) but less stained in the kidney of DM⁺RAGE (−/−) and nondiabetic animals [DM^-RAGE (+/+) and DM^-RAGE (−/−)] (Fig. 2J), suggesting that local S100-RAGE interactions might also affect the development of diabetic nephropathy. However, serum S100 levels detected by immunoblotting were not changed among the four groups: DM^-RAGE (+/+) 100 ± 15.2%; DM^-RAGE (−/−) 84.8 ± 23.1%; DM⁺RAGE (+/+) 93.1 ± 20.3%; and DM⁺RAGE (−/−) 102.2 ± 28.4% (data not shown). These findings suggested that the deletion of RAGE prevented the development of diabetic kidney changes.

To overcome diabetic nephropathy, RAGE would thus be a candidate molecule worth targeting. However, no RAGE antagonists were available. Since RAGE is a heparin-binding protein, we speculated that heparin might influence AGE-RAGE interactions. We initially tested the unfractionated high–molecular weight heparin. It was found to inhibit AGE association with RAGE. Its net action on RAGE, however, was agonistic when assessed with human endothelial cells in culture (data not shown). We next examined whether fragmented and fractionated heparin, LMWH, can inhibit the AGE-RAGE interaction and antagonize RAGE signaling. LMWH largely lost N-sulfate residues during the fragmentation by nitrous acid and its biological actions were speculated to be different from unfractionated high–molecular weight heparin. First, we used SPR assay to check LMWH binding to the purified extracellular domain of RAGE. The assay revealed a direct association with the K_d of 17 nmol/l (Fig. 4A). Secondly, a dose-dependent competitive inhibition by LMWH was demonstrated in a plate assay of AGE binding to RAGE (Fig. 4B). Thirdly, to check its antagonistic activity in RAGE signaling, we performed a cellular NFκB promoter assay using RAGE-overexpressing cells. Increased NFκB promoter-driven luciferase activity by AGE-BSA was significantly blocked by the addition of LMWH (Fig. 4C). RAGE dependency and specificity of the NFκB activation were clearly defined by using soluble RAGE (sRAGE), RAGE siRNA, and a dominant-negative RAGE lacking the intracellular domain. Finally, 0.1–1.0 IU/ml of LMWH have significantly inhibited AGE-induced VEGF mRNA upregulation in RAGE-overexpressing ECV304 cells and AGE-induced VCAM-1 expression in HUVEC (Fig. 4D and E).

We then conducted an interventional study with LMWH. Mice were divided into five groups: one nondiabetic control group and four iNOSTg diabetic groups that received LMWH infusion via an osmotic pump and daily subcutaneous injection of 40 or 80 IU of LMWH or vehicle alone. Each group was treated and observed from 1 month (subclinical state of diabetic nephropathy) to 4 months (overt and advanced nephropathy state) of age. Pharmacokinetic study of LMWH in the dose of 40 and 80 IU s.c. groups showed that the maximum plasma concentrations of LMWH were 0.10 and 0.15 IU/ml, respectively, at 1 h after injection (Fig. 5A). A weekly check showed that plasma LMWH concentrations were 0.10–0.15 and 0.05–0.10 IU/ml in the LMWH 80 and 40 IU daily injection groups, respectively, and 0.05–0.10 IU/ml in the LMWH osmotic pump group during the 3-month observation period (Fig. 5B). LMWH treatment did not lower the blood glucose and A1C levels of each group (Fig. 5C and D). Periodic acid Schiff (PAS) staining of the kidneys showed reduced deposition of PAS-positive materials in the glomeruli of the LMWH-treated groups (Fig. 5F). By quantitative evaluations, LMWH dose dependently and significantly reduced urinary albumin creatinine ratio (Fig. 5E), glomerular tuft area (Fig. 5G), glomerular cell number (Fig. 5H), and glomerular sclerosis index (Fig. 5I) when compared with the control diabetic group. Finally, we performed a therapeutic study with LMWH in diabetic mice from 4 (advanced diabetic nephropathy state confirmed by open renal biopsy) to 5 months of age. Treatment with LMWH 80 IU daily s.c. injection for 1 month revealed significant reduction in urinary albumin creatinine ratio, mesangial expansion, and glomerulosclerosis but not in glomerular cell count (Fig. 6A–D). Blood glucose and A1C levels were again not affected. LMWH 80 IU daily s.c. treatment for 1 week did not affect urinary albumin creatinine ratio in diabetic RAGE-null mutants (Fig. 6E).

We examined the expression of VEGF, TGF-β, and CTGF proteins in the kidney by immunostaining; these factors are well known to link to diabetic nephropathy. The VEGF was faintly expressed, but detected, in the glomeruli from DM⁺RAGE (+/+) (Fig. 7A). The TGF-β and CTGF protein expression levels were the highest in DM⁺RAGE (+/+) compared with DM⁺RAGE (−/−) and with nondiabetic animals [DM^-RAGE (+/+) and DM^-RAGE (−/−)] (Fig. 7B and C). LMWH treatment suppressed the expressions of TGF-β and CTGF protein in the glomeruli (Fig. 7).

Diabetic nephropathy is a chronic and severe complication of diabetes, causing increased mortality due to ESRD. After the onset of diabetic nephropathy, albuminuria, glomerular hypertrophy, and nephromegaly appear in early phase, followed by mesangial expansion, glomerulosclerosis, and increased serum creatinine in advanced phase (32). The iNOSTg diabetic mice used in this study developed kidney complications, similar to those seen in humans, when maintained on a CD-1 background and fed a high-calorie diet. They exhibited albuminuria, nephromegaly, and increased expressions of VEGF, TGF-β, and CTGF until 4 months of age; thereafter, they displayed progressive mesangial expansion, glomerulosclerosis, and increased serum creatinine levels (Figs. 2, 3, and 7). As consistent with our previous results that showed an acceleration of diabetic nephropathy in RAGE-overexpressing mice, the disruption of the RAGE gene was found in this study to ameliorate all of the early- and advanced-phase indexes of diabetic nephropathy (Figs. 2 and 3). Moreover, the extent of this attenuation was proportional to the RAGE gene dosage (Figs. 2 and 3). Recently, administration of sRAGE or of neutralizing RAGE antibody was reported to block the early events of diabetic kidney changes in db/db mice, an obese type 2 diabetic mouse model that develops only early-stage nephropathy (9,33). Moreover, it was previously reported that streptozotocin-induced diabetic RAGE-null mutant mice did not display early events in the kidney such as nephromegaly, increased expression of VEGF, and TGF-β in the renal cortex (9,34). Together with these findings, our data indicate that RAGE contributes to the full expression of diabetic nephropathy and should be a target for disease intervention. Circulating levels of AGE were essentially invariant among three diabetic groups, whereas AGE deposition in renal glomeruli was greater in the DM⁺RAGE (+/+) mice than in the DM⁺RAGE (−/−) mice (Fig. 2I). This finding suggests that AGE-RAGE interactions in situ would initiate and superdrive the development of renal lesions. Unexpectedly, serum AGE levels in the DM^-RAGE (−/−) mice were high similar to those observed in the diabetic groups (Table 1) but lacking deposition in the glomeruli (Fig. 2I). This result suggests that in addition to being a signaling receptor, RAGE might function as a clearance receptor for circulating AGE molecules.

RAGE would seem, therefore, to be a promising target for overcoming diabetic nephropathy, and if a compound was available to antagonize the AGE action on RAGE, an effective remedy against this disease could be developed. In this study, an SPR assay revealed that LMWH is able to bind RAGE with an affinity approximately six times higher than AGE (i.e., ∼17 vs. ∼100 nmol/l) (Fig. 4). The mode of action of LMWH was antagonistic as shown by the competition of AGE in RAGE binding experiments and by the inhibition of RAGE-dependent NFκB activation in glioma cells and of AGE-induced VEGF and VCAM-1 expression in endothelial cells (Fig. 4C–E). LMWH blocked amphoterin/HMGB1-RAGE signaling by LMWH binding to both amphoterin/HMGB1 and RAGE but did not block S100B-RAGE signaling (data not shown). This suggests that the S100-binding site may be different from the AGE-binding region of RAGE, or LMWH may not completely block the S100 binding to RAGE. When LMWH was administered to the diabetic nephropathy mouse model, both preventive and therapeutic effects were noted by the evaluation of the representative indexes of diabetic nephropathy such as albuminuria, mesangial expansion, and glomerulosclerosis (Figs. 5 and 6). These in vivo effects were achieved with 0.1–0.15 IU/ml LMWH within the therapeutic range of plasma LMWH concentrations. That is to say that therapeutic LMWH concentrations do not cause prolongation of activated clotting time or serious side effects (35,36). Consistent with our findings, it was reported that oral sulodexide, a mixture of glycosaminoglycans (GAGs) composed of 80% LMWH and 20% dermatan sulfate improved albuminuria in both type 1 and type 2 diabetic patients (37). The long-term administration of exogenous GAGs is also reported to have favorable effects on both morphologic and functional renal abnormalities and glomerular hemodynamics in diabetic nephropathy (38). The anticoagulant effects of LMWH were speculated to contribute to the improvement of diabetic nephropathy by affecting the microcirculation. Several studies have provided evidence that heparan sulfate and heparin may interfere with signaling events (39). In addition, the action of a variety of growth factors (e.g., fibroblast growth factor and VEGF) is known to be regulated by binding to heparan sulfate on the cell surface or the extracellular matrix, which promotes or restricts interactions with their signal transducing receptors. In vascular endothelial cell culture, LMWH is reported to prevent hyperglycemia-induced endothelial dysfunctions such as expressions of intercellular adhesion molecule-1, VCAM-1, and E-selectin and translocation of NFκB, as well as generation of intracellular reactive oxygen species (40). However, little is known about the evidence and mechanism by which LMWH influences the cellular events as the heparan sulfate or heparin does. This study suggested that the antagonistic effect of LMWH on RAGE at least partly contributed to the improvement of diabetic nephropathy. Immunohistochemically reduced S100 protein staining was observed in the LMWH-treated diabetic mice compared with control diabetic mice (data not shown); this may be suggesting another action of LMWH. The protective effects of LMWH and GAGs have been ascribed to the inhibition of the diabetes-induced upregulation of renal TGF-β (41). We observed that LMWH treatment could inhibit AGE-induced TGF-β upregulation in cultured mesangial cells from wild-type mice but not in those from RAGE-null mice (data not shown). As LMWH is also known to be cleared principally by the renal route (42), it may help facilitate the renoprotection from diabetic insult.

In conclusion, this study firmly established RAGE as a promising target, and RAGE antagonists will be a useful remedy for both prophylaxis and treatment of diabetic nephropathy.

This work was supported in part by the “Research for the Future” Program of the Japan Society for the Promotion of Science (grant 97L00805); by grants-in-aid from the Ministry of Education, Science, Sports, Culture and Technology, Japan (no. 13670113) and for scientific research from the Japan Society for the Promotion of Sciences (nos. 16790183, 17590241, and 16570113); from the Japan Diabetes Foundation; and from the Sankyo Foundation of Life Science.

We thank S. Matsudaira and R. Kitamura (Kanazawa University) and H. Hatta (University of Toyama) for their assistance.