Activation of Tubular Epithelial Cells in Diabetic Nephropathy

Reagents were obtained as follows: HEPES buffer solution, l-glutamine, penicillin-streptomycin mixture, and PBS, pH 7.4, were obtained from Biowhittaker (Walkerville, MD). pTEC medium and IL-6 ELISA were obtained from Promocell (Heidelberg, Germany). FCS was from Gibco/BRL (Dreieich, Germany). [γ-³²P]ATPα (3,000 Ci/mmol at 10 Ci/ml), Hybond-N-Nylonfilter, ECL-nitrocellulose membranes, ECL detection reagents, and Hyperfilm X-ray films were obtained from Amersham (Braunschweig, Germany). PMSF, thioctic acid (TA), and the Limulus assay were purchased from Sigma (Deisenhofen, Germany). Poly dI/dC was from Pharmacia (Freiburg, Germany). Polyclonal anti-RAGE antibodies, generated in goat with recombinant RAGE prepared in E. coli as antigen, were a gift from Dr. M.A. Shearman (Merck, Sharpe & Dome, Essex, U.K.). Vectastain ABC kit was purchased from Vector Laboratories (Burlingame, CA). Anti-p65, -p50, -p53, -cREL, -relB, and -κβα and the respective second antibodies were obtained from Santa Cruz (Heidelberg, Germany). Monoclonal anti-p65 antibodies specific for activated NF-κBp65 and Fugen 6 transfection reagent were obtained from Roche (Mannheim, Germany). Soluble RAGE (sRAGE) preparations used throughout this study have previously been described in detail (70,71,83,92,96) and were generously provided by Drs. Schmidt and Stern (Columbia University, New York). The kit for the determination of CML antigen was kindly provided by Rosemarie Kientsch-Engels (Roche AG, Penzberg, Germany).

Kidneys obtained from diabetic BB/O(ttawa)K(arlsburg) rats were used as diabetes model. This animal model is described in detail by Klöting et al. (98). Diabetes was present for 57 ± 9 days at an age of 104 ± 16 days. Rats were treated with a continuous infusion of insulin at 2 units/24 h, and the blood glucose level was kept at >20 mmol/l. Kidneys from normal Sprague-Dawley rats served as a control. Paraffin-embedded kryostat sections from rat kidneys were prepared from normal and diabetic animals according to a standard protocol.

Immunohistochemistry was performed on paraffin-embedded tissues of a human kidney (obtained from kidney biopsy) and rat kidneys by indirect immunoperoxidase technique. Detection of signals was performed with the Vectastain ABC kit (Vector Laboratories) according to the manufacturer’s instructions as described (68–70,75,99–101). Peroxidase activity was visualized with 0.05% 3,3-diaminobenzidine-tetrahydrochloride (Serva, Heidelberg, Germany) before the sections were counterstained with Mayer’s hematoxylin. Controls for immunstaining were included in all stainings by omission of the primary antibody and its replacement by PBS and matching concentrations of normal rabbit serum (data not shown). In addition, blocking peptides were included in some of the reactions to confirm specificity (data not shown).

For the investigations, each patient gave informed consent and the study was approved by the ethical committee of the Department of Medicine, University of Heidelberg, and performed in accordance with the Declaration of Helsinki.

Urine samples from healthy control subjects (n = 50) and type 2 diabetic patients (n = 100) were collected from morning spot urine and immediately stored at −20°C. Information about the clinical and laboratory data from the patients is given in Table 1.

Patients with nondiabetic kidney disease were suffering from glomerulonephritis, systemic lupus erythematodes, or vasculitis.

Quantification of urinary CML was performed using a commercially available kit as previously described (102).

Immunocytochemistry staining was performed as described in detail elsewhere (70,79,87). In brief, fresh urine samples were obtained from healthy control subjects (n = 50) and type 2 diabetic patients (n = 100). After direct centrifugation, the supernatant was decanted and the concentrated cellular material was deposited on glass slides by cytocentrifugation (Shandon Cytospin). Tubular epithelial cells were identified on May-Grünwald/Giemsa (Pappenheim) staining and by positive reactions with antibodies against cytokeratin 18 and neutral endopeptidase. In a second step, cells were stained with a monoclonal mouse antibody against NF-κBp65 antigen (Roche) and a FITC-labeled monoclonal mouse antibody against IL-6 antigen using a standard protocol. The NF-κB antibody recognizes activated NF-κBp65 (103). Staining was performed using anti-NF-κBp65 and anti-IL-6 at a concentration of 0.01 μg/ml for 60 min at room temperature.

Human renal pTECs were obtained from adult kidneys after surgery from the unaffected parts of kidneys obtained from tumornephrectomy or kidney biopsies, as described (104–106). Proximal tubular epithelial cells (pTECs) were characterized by staining with FITC-labeled cytokeratin 18 antibodies and an antibody against neutral endpeptidase as described in immunocytochemistry. For electrophoretic mobility shift assay (EMSA) analysis, Western blot, and enzyme-linked immunosorbent assay (ELISA), pTECs from the same passage were used (fourth to fifth passage). Before stimulating pTECs with AGE-albumin or CML, cells were cultivated without growth factors for 5 days. Where indicated, cells were preincubated with sRAGE (a soluble and truncated form of the receptor RAGE) (70,71,75,83,92,93,96) in a threefold molar excess (1.5 μmol/l) compared with AGE-albumin or CML, a RAGE-specific antibody (20 ng/μl), or TA (200 μmol/l).

AGE-albumin was prepared as previously described (70,96,107). The extent of lysine modifications in the AGE preparations varied up to 36%. In vitro synthesis of CML-albumin was performed as previously described by Schleicher and colleagues (52,96). Assays for endotoxin showed AGE-albumin and CML preparations to contain virtually undetectable levels of lipopolysaccharide (LPS) (<10 pg/ml at a protein concentration of 5 mg/ml according to the Limulus assay [Sigma]).

After stimulation of pTECs with AGE-albumin and CML with the concentrations and time points indicated in the figure legends (Figs. 3, 4, and 6), nuclear proteins from pTECs were harvested as described elsewhere (67,70,96,99,100) and assayed for transcription factor binding activity using the NF-κBp65 consensus sequence: 5′-AGTTGAGGGGACTTTCCCAGGC-3′. Specificity of binding was ascertained by competition with a 160-fold molar excess of unlabeled consensus oligonucleotides and supershift experiments. For supershift experiments, nuclear extracts were preincubated with antibodies against NF-κBp65, -p50, -p52, -cRel, and -relB prior to stimulation as described elsewhere (50,51,56,67,69,70,96,99,100,108–111). All experiments were performed at least three times.

Cytoplasmic and nuclear fractions were prepared as previously described in detail (69,70,99,100). Western blot was performed as described. Membranes were incubated with primary antibodies directed against NF-κB65 and -κβα. After washing, the secondary antibody (horse radish peroxidase-coupled rabbit IgG) was added and incubation was continued for 30 min. Immunoreactive proteins were detected with the ECL-Western blot System (Amersham Pharmacia, Braunschweig, Germany) and subsequent autoradiography for 2 min. All experiments were performed three times.

The simian virus 40–driven luciferase control plasmid pGL2-control, the promoterless plasmid “pGL2-basic,” and the β-galactosidase control plasmid “pSV-Gal” were obtained from Promega (Heidelberg, Germany). The plasmid NF-κB-Luc, which contains four tandem copies of the NF-κB consensus sequence fused to a TATA-like promoter region from the Herpes simplex virus thymidine kinase promoter, was purchased from Clontech (Heidelberg, Germany). The Iκβα expression plasmid was kindly provided by Dr. Baeuerle (Tularic Inc.).

For transfection experiments, pTECs growing in the logarithmic phase were transfected as described (67,70,93,96,99,100,112) using Fugen 6 transfection reagent (Roche) according to the manufacturer’s instructions. Before cells were stimulated with AGE-albumin or CML (concentrations and time points are indicated in the figure legends), medium was changed and cells were kept without growth factors and FCS. Cotransfection was performed as previously described (69,70,96,100) with a κBα-overexpressing plasmid. After 42 h, cells were washed with 37°C warm NaCl 0.9% for two times and harvested as described elsewhere (69,70,96). For inhibition experiments, TA (200 μmol/l), sRAGE (1,500 nmol/l), or anti-RAGE (20 ng/μl) was combined with AGE-albumin or CML (500 nmol/l). The ratio of luciferase activity to β-galactosidase activity served to normalize luciferase activity (112). Each experiment was performed in triplicates, and experiments were repeated at least three times.

The supernatant from NF-κBp65 or inhibitory κB (IκB)α-transfected and AGE- or CML-stimulated cells was harvested and IL-6 antigen determined by ELISA. The ELISA for determination of IL-6 was performed using a commercially available kit (Promocell) according to the instructions of the manufacturer. The experiment was performed in triplicates and repeated at least three times.

All values are given as mean, with the bars showing SDs. For statistical analysis, Student’s t test, Fisher’s test, Mann-Whitney U test, Pearson correlation, χ² median test, and determination of contingency coefficient (CC) were performed. P < 0.05 was considered statistically significant.

Paraffin-embedded tissues from diabetic BB/O(ttawa)K(arlsburg) (98) rat kidneys were prepared to demonstrate NF-κBp65 and -p50 antigen. Diabetes was present for 57 ± 9 days and manifested at an age of 104 ± 16 days. Normal Sprague-Dawley rats served as a control group. In diabetic animals, NF-κBp50 and -p65 antigen was present in the nuclear region of tubular cells (Fig. 1B and D), while in control animals, no NF-κBp50 and -p65 antigen could be detected in tubular cells (Fig. 1A and C). Interestingly, in diabetic kidneys, NF-κBp50 and -p65 was mainly present in tubular cells and not in glomerular cells. To confirm these data, kidney specimens derived from a kidney biopsy of a patient with diabetic nephropathy were stained with the same antibodies. Tubular epithelial cells showed a marked staining for NF-κBp65 and -p50 antigen (Fig. 1E and F). Again, NF-κBp65 and -p50 were mainly present in the renal tubular system.

Urine was collected from 50 healthy control subjects and 100 patients with type 2 diabetes (50 patients with normal albumin excretion [<20 mg/l] and 50 with macroalbuminuria [>200 mg/l]). Mean serum creatinine and urea had been normal. In addition, urine from 50 nondiabetic patients with macroalbuminuria [>200 mg/l] was collected. Patient characteristics are shown in Table 1. Urinary excretion of CML-modified proteins was determined by ELISA. Diabetic patients showed significantly higher urinary CML levels than healthy control subjects (P < 0.0001) or nondiabetic patients with macroalbuminuria >200 mg/l (P < 0.0001). These results are in contrast to a recently published study that showed a decreased urinary excretion of CML in diabetic patients with impaired renal function (102). This is most probably due to an increase in CML excretion, while renal function is not severely impaired. With decreasing renal function, CML can be excreted only to a lesser extent. When the diabetic patients were further analyzed, a positive correlation between albuminuria and excretion of CML antigen (r = 0.7, P < 0.0001) was found, while there was no correlation between CML excretion and HbA_1c (r = 0.03, P = 0.76).

The presence of pTECs in the urine was evidenced using antibodies against cytokeratin 18 and neutral endopeptidase. Single or scattered pTECs (Fig. 2A and B) were found in the urine of 20 of 100 type 2 diabetic patients but in none of the healthy control subjects (P < 0.0001). Positive staining for activated NF-κBp65 antigen was recognized in 8 of 20 (40%) patients, in some but not all excreted pTECs (Fig. 2C). Furthermore, to indirectly assess the transcriptional consequences of NF-κBp65 activation, slides were incubated with an antibody against IL-6. In five of eight patients positive for NF-kBp65, IL-6 antigen could be detected (62%) in single cells (Fig. 2D), whereas in the preparations negative for NF-κBp65, only one expressed the IL-6 antigen (1/12).

pTEC positivity in the urine sediment showed a strong correlation to the degree of albuminuria (χ² median test = 25, CC = 0.63) and CML excretion (χ² median test = 12.5, CC = 0.46 (113), leading to the hypothesis that CML-modified proteins may induce NF-κBp65 positivity in pTECs.

In vitro experiments were performed using cultivated human pTECs to prove that tubular cells indeed have the ability to activate NF-κB in response to increased AGEs. pTECs were stimulated with either AGE-albumin or CML as described in research design and methods. When cultured pTECs were incubated over 6 h with either AGE-albumin (Fig. 3A) or CML (Fig. 3B), a dose-dependent activation of NF-κB was observed in EMSA (Fig. 3A and B). NF-κB binding activity was half maximal at 250 nmol/l AGE-albumin and 500 nmol/l CML. Normal nonglycated human albumin did not induce NF-κBp65 (Fig. 3E). Heat inactivation of AGE-albumin over 12 h at 100°C abolished inducible NF-κB binding activity (data not shown).

AGE-albumin and CML induced NF-κB binding activity in a time-dependent manner (Fig. 3C and D). An early start of NF-κB activation was observed already at 30 min, reaching a first maximum after 6 h (Fig. 3C, lane 4). After a decrease between 12 and 24 h (Fig. 3C, lanes 5 and 6), a second peak could be observed after 4 days (Fig. 3C, lane 7). This time course is similar to the data obtained in a previous study (70). Stimulation of pTECs with CML revealed similar data compared with stimulation with AGE-albumin (Fig. 3D).

Supershift analysis (Fig. 4) revealed that NF-κBp65 and -p50 (lanes 4 and 6) constituted the major protein contributing to the shift observed, while NF-κBp52, -cRel, and -RelB did not participate in the binding reaction. NF-κB binding activity was suppressed using a sixfold excess of unlabeled oligonucleotides (lane 9). Heat-inactivated AGE did not activate NF-κB in cultured pTECs (lane 1).

Western blot analysis corresponded well to the binding activity demonstrated in EMSA analysis. After AGE-albumin or CML stimulation, NF-κBp65 antigen translocation was dose dependent (Fig. 5A and B). The decrease in cytoplasmic p65 antigen occurred after 60 min, simultaneously with the increase in nuclear p65 antigen (Fig. 5C and D). Correspondingly, IκBα degradation was also dose and time dependent (Fig. 5A–D). A reconstitution of cytoplasmic NF-κBp65 antigen was observed at 3–4 days (lanes 8 and 9), a time point of maximal NF-κBp65 binding activity (Fig. 3) and nuclear translocation (Fig. 5C and D). A previous report demonstrated that long-lasting NF-κBp65 activation is associated with increased NF-kB synthesis, overriding IκBα, explaining simultaneous nuclear and cytoplasmic NF-κBp65 antigen (70). Consistently, a strong loss of κβ α antigen was observed after a 12-h stimulation, but not after 4 days (Fig. 5C and D), presumably because NF-κBp65 drives the de novo synthesis of IκBα (70).

As shown by EMSA (Fig. 3), NF-κBp65 activation started after 30 min, reaching a first maximum after 6 h and a second maximum at 4 days after CML stimulation and thus resembled the activation pattern observed for AGE-albumin. In Western blot analysis, however, translocation of NF-κBp65 from the cytoplasm into the nucleus was already observed 30 min after AGE stimulation, but not 6 h after CML stimulation, which might be due to a lesser stimulatory effect of CML compared with AGE-albumin. Since EMSAs are much more sensitive than Western blots, it is reasonable to assume that a weaker CML-dependent NF-κB induction can be monitored in EMSA but is not evident in Western blot analysis.

All effects could be reduced using the antioxidant TA. The reduction could be observed in EMSA analysis and Western blot (Fig. 6A and B and Fig. 7A). To investigate whether the AGE-albumin- and CML-induced NF-κBp65 activation is RAGE dependent, coincubation with sRAGE and a specific RAGE-antibody was performed. sRAGE and RAGE antibody decreased NF-κBp65 activation in EMSA (Fig. 6C and D) and Western blot (Fig. 7B).

Transient transfection of cultured pTECs, using a NF-κB consensus-driven luciferase reporter plasmid, was performed to demonstrate that increased NF-κB binding activity (Fig. 3) and nuclear translocation (Fig. 5) is functionally significant and results in increased NF-κB-dependent gene expression. A dose-dependent activation could be demonstrated when pTECs were stimulated with AGE-albumin or CML (Fig. 8A and B), corresponding well to the data concerning NF-κB activation in EMSA analysis (Fig. 3) and Western blot (Fig. 5). When cells were stimulated for 6 h with 500 nmol/l AGE-albumin or CML, respectively, an 8- (CML) to 10-fold increase (AGE) in luciferase activity was observed (Fig. 8A–D). This was not the case when cells were stimulated with control albumin. Overexpression of κβ and treatment with the antioxidant TA reduced luciferase activity (Fig. 8C and D). Furthermore, NF-κB activation is RAGE dependent, since AGE-albumin-induced binding activity was markedly inhibited by addition of excess sRAGE and RAGE antibody, as described in research design and methods (Fig. 8C and D). Due to experimental limitations of transient transfection experiments, the time of stimulation could not be exceeded for >42 h.

AGE-albumin- and CML-mediated pTEC activation resulted not only in increased expression of luciferase but also in de novo synthesis of IL-6 antigen (studied as a model of a NF-κB-driven gene). When the supernatants of the transfected cells (transfection data are shown in Fig. 8) were analyzed for IL-6 antigen, a dose-dependent increase was observed (Fig. 9A and B). IL-6 antigen release was reduced by overexpression of IκB or addition of the antioxidant TA. Furthermore, the IL-6 antigen induction is RAGE dependent, since sRAGE and RAGE antibody reduced IL-6 release (Fig. 9C and D).

Human pTECs are not only passive bystanders in the development of diabetic nephropathy, but they also respond actively to hyperglycemia and AGEs by inducing NF-κB activation and NF-κB-dependent gene expression in vitro and in vivo. One defined AGE generated by lipoxidation and glycoxidation in diabetic nephropathy is CML (44,52,61,70,114). The presence of CML-modified proteins in the urine of type 2 diabetic patients and the in vitro observation that CML is a potent inducer of sustained NF-κB activation in pTECs suggest that CML might play a role in the development of diabetes renal complications. In addition, the observation that type 2 diabetic patients demonstrated excretion of tubular cells that was positive for activated NF-κBp65 and IL-6 antigen implies that the AGE/CML-RAGE-mediated NF-κB activation is functionally significant.

Indirect evidence for the role of NF-κB activation in diabetic nephropathy has already been given from clinical studies in which an increase in oxidative stress correlated with renal function and NF-κB activation in patients with type 2 diabetes (99,100,115–118). We found that increased CML excretion in type 2 diabetes correlates to the excretion of NF-κBp65 and IL-6 antigen-positive pTECs.

As demonstrated here, pTECs respond to exogenously added AGE-albumin and CML with an NF-κB activation that meets the requirements of RAGE-dependent NF-κB activation, as evidenced by NF-κB binding activity that lasted >4 days and NF-κB-dependent gene expression (69,70). In vivo, pTECs are exposed not only to AGEs present in the urine but also to glucose leading directly to intracellular AGE formation (41). An additional source of intracellular CML formation is the inflammatory reaction, as we have shown that pTECs in diabetic patients are in part positive for IL-6. Cytokines and inflammatory agents, as they occur in time remodeling, are associated with intracellular CML formation and activation of NF-κBp65. It remains unknown whether pTECs are subject to extracellular- and/or intracellular-mediated CML responses in diabetes. Until now, only cell responses to extracellular AGEs, for example via RAGE, have been reported. The involvement of RAGE in AGE-dependent pTEC activation was confirmed by competition of NF-κB activation by sRAGE and an RAGE blocking antibody. This in agreement with recent data showing reduction of vascular hyperpermeability by scavenging RAGE ligands (75,91,92,119,120) and increased diabetic nephropathy in diabetic animals overexpressing RAGE (93). Furthermore, there is evidence that RAGE-ligand interaction contributes to sustained NF-κBp65 activation (41,68–70,83,121,122). The long-lasting nature of RAGE-dependent NF-κBp65 activation corresponds well to the p65 and p50 positivity of tubular cells in the tissue sections shown in Fig. 1. One would not expect all cells to be positive, even in serial sections, if NF-κB activation in humans would be as short lasting as in tissue culture after tumor necrosis factor stimulation. Thus, in diabetes, autoregulatory negative feedback loops are shut down. As shown previously, one mechanism is excessive de novo synthesis of p65, overriding IκBα inhibition (70,123–125). Since many stimuli result in NF-κB activation, one has to assume that mechanisms for cell- and disease-specific activation must exist.

It has been demonstrated that different stimuli in different renal diseases lead to disease-specific activation of certain NF-κB subunits. Using LPS as stimulus, or other renal models like rats with ureteric obstruction or immune complex nephritis, different activation patterns could be observed (50,51,55,56,108–111,126). This suggests that differences in the NF-κB activation pattern in response to a given stimulus might determine the selection of the genes activated. Therefore, we investigated which activation pattern is present in human pTECs due to AGE-albumin stimulation. We and others show two specific DNA-protein complexes (50,51,56,111). Supershift analysis revealed that the slower-migrating complex is composed of NF-κBp65 and -p50. In accordance with other studies, the faster-migrating band could not be depleted could by NF-κBp65, -p50, -p52, -cREL, or -relB antibodies but could by unlabeled oligonucleotide. To define whether the complex is due to the binding of a coactivator protein such as CBP/p300 (111), however, is beyond the scope of this study.

The observation that excreted pTECs demonstrate both NF-κB and IL-6 antigen activation led us to speculate that a perpetuated increase in proinflammtory gene products such as IL-6, depending on perpetuated NF-κB activation (as demonstrated in vitro), might be central in damaging the renal tubule. This view is emphasized by the fact that not only cytokines, but also metalloproteinases such as MMP9, are controlled by NF-κB. MMP-9 has been implicated to contribute to proteinuria in Heymann nephritis and therefore might be a good candidate for the destruction of pTECs in the course of diabetic nephropathy (127).

Further studies are needed to define which mechanisms are activated by perpetuated NF-κB activation and finally result in the destruction of the tubulus, as evidenced by the excretion of pTECs in overt diabetic nephropathy.

This work was supported by grants from the University of Heidelberg (to M.M. and A.B.) and the University of Tübingen (to A.B. and P.P.N.), the IZKF program of the University of Tübingen (to P.P.N.), and Asta Medica (to A.B. and P.P.N.). P.P.N. performed part of this work during the tenure of a Schilling professorship. A.B., B.Y., E.S., H.U.H., F.v.W., and P.P.N. were supported by the Deutsche Forschungsgemeinschaft. A.-M.S. and D.S. were supported by grants from the USPHS, the American Heart Association (New York affiliate), and the Juvenile Diabetes Foundation. A.A.R.S. is supported by a grant from the Arab Republic of Egypt.

We thank Dr. M.A. Shearman (Merck, Sharpe & Dome) for providing anti-RAGE antibodies, Dr. Baeuerle (Tularic Inc.) for providing Iκβα expression plasmid, and Dr. Kientsch-Engels (Roche AG) for providing the CML-ELISA kit. Part of this report was presented at the meeting of the German Diabetes Society 2001 (Aachen, Germany).