Damage to the Peritoneal Membrane by Glucose Degradation Products Is Mediated by the Receptor for Advanced Glycation End-Products

Abstract

Peritoneal dialysis is limited by morphologic changes of the peritoneal membrane. Use of peritoneal dialysis fluids (PDF) that contain glucose degradation products (GDP) generates advanced glycation end-products (AGE) within the peritoneal cavity. It is unknown whether peritoneal damage is causally related to AGE–receptor for AGE (RAGE) interaction. The effects of PDF were compared with different amounts of GDP on morphologic changes of the peritoneal membrane in 48 wild-type (WT) and 48 RAGE-deficient mice. PDF (1 ml) were instilled twice daily over a period of 12 wk. Groups with eight animals each received no manipulation (sham); sham instillation (sham i.p.); or filter-sterilized, glucose-free, conventional low GDP- or high GDP PDF. In vitro (generation of AGE fluorescence in PDF) and in vivo (immunohistochemistry for carboxymethyllysine), a GDP-dependent increase of AGE formation occurred. Inflammation and neoangiogenesis were augmented in WT mice that were treated with high GDP accompanied by upregulation of CD3⁺ T cells, increased NF-κB binding activity, increased lectin, and vascular endothelial growth factor expression. Furthermore, pronounced submesothelial fibrosis was found with increased expression of TGF-β1. Exposure to low GDP resulted in only mild inflammation and neoangiogenesis (compared with sham i.p.) and no fibrosis in WT mice. The findings in WT contrasted with those in RAGE-deficient mice, which showed no increased inflammation (CD3⁺ T cells and NF-κB binding activity), neoangiogenesis (by lectin and vascular endothelial growth factor expression), or fibrosis (expression of TGF-β1) after long-term exposure to GDP-containing PDF. Peritoneal damage by GDP in PDF is dependent at least in part on AGE–RAGE interaction.

Peritoneal dialysis (PD) is an accepted alternative to hemodialysis in the treatment of patients with ESRD. Long-term PD is limited by structural and functional changes of the peritoneal membrane resulting in dialysis failure (¹). Conventional PD fluids (PDF) are a source of glucose degradation products (GDP; carbonyl compounds) that are generated during heat sterilization under acidic conditions (²). Such GDP are highly reactive substances (^2–10) that exhibit considerable direct cytotoxicity. However, GDP lead to the formation of advanced glycation end-products (AGE) (¹¹). The extent to which peritoneal damage from GDP is due either to direct effects, e.g., illustrated by generation of vascular endothelial growth factor (VEGF) by methylglyoxal in peritoneal cells (¹²), or to indirect effects such as the generation of AGE from precursor GDP is unknown. To investigate mechanisms of AGE-dependent peritoneal damaging, we compared PDF with different GDP contents in wild-type (WT) and receptor for AGE (RAGE)-deficient (−/−) mice. The rationale for this approach was based on the consideration that RAGE, a multiligand member of the Ig superfamily (^13,14), is the best characterized signal transduction receptor for AGE. Binding of AGE to RAGE results in an activation of key signal transduction pathways, such as NF-κB and subsequent transcription of mediators for which a role was also claimed in the pathogenesis of uremic complications (^15,16). To avoid this confounder, we performed the following study in nonuremic animals.

Human peritoneal mesothelial cells also express RAGE (¹⁷). Furthermore, monoclonal anti-RAGE antibodies prevent fibrosis of the peritoneal membrane induced by hyperglycemia in a diabetic animal model (¹⁸). However, it is not certain whether anti-RAGE antibodies recognize structures other than the ligand binding domain of RAGE. The purpose of our experimental study was to address the following issues: (1) Whether AGE–RAGE interruption, demonstrated in a RAGE −/− mouse model, prevents peritoneal damage after long-term PD; (2) whether interruption of AGE–RAGE interaction also prevents peritoneal damage in a nondiabetic model; and (3) whether the adverse effects of GDP (as opposed to glucose per se) are mediated via RAGE.

Materials and Methods

Animal Experiments

The experiments were performed in female mice (48 WT mice expressing RAGE, genetic background C57BL/6 and 48 RAGE −/− mice) that weighed 31.6 ± 0.54 g according to the guidelines of the Institute of Laboratory Animal Science of the University of Heidelberg. The RAGE-targeting construct and the generation of RAGE −/− mice have been described in detail elsewhere (¹⁹). The abrogation of RAGE expression at the mRNA level was analyzed by reverse transcription–PCR of lung tissue, in which RAGE is constitutively expressed at high levels. Pair-fed mice were kept in a 12-h day-night rhythm at a constant room temperature of 20°C. WT mice and RAGE −/− mice were randomly allocated to six groups, as shown in Table 1.

Twice daily, the animals received an intraperitoneal (i.p.) injection of 1 ml of the respective PDF solution at 37°C under sterile conditions. The composition of the PDF is given in Table 2. One group of mice was kept completely unmanipulated to obtain baseline histologic and molecular data (sham). In a second group, mice were sham-injected without instilling solution as a control for puncture trauma (sham i.p.). The experiment was terminated for all animals after 12 wk. We also considered the possibility of inflammatory injury to the peritoneum by a reaction to the trauma of twice-daily injection. Indeed, there was some reaction to injection, as evidenced by comparison of sham animals and sham i.p. animals, but the degree of reaction was significantly less than in response to the dialysis fluid. We took particular care to avoid artificial injury by thermal trauma (fluids were injected at body temperature) and contamination by bacteria or lipopolysaccharides.

Histologic and Immunohistochemical Examinations

Visceral peritoneal tissue samples were fixed in 6% phosphate-buffered formalin (pH 7.4) and embedded in paraffin. Four-micrometer-thick tissue sections were stained with hematoxylin & eosin, periodic acid Schiff reagent, and Picro Sirius red staining for detection of fibrous tissue and investigated by light microscopy.

Immunohistochemical stainings were performed in three groups (high GDP, low GDP, and sham i.p.). Tissue sections were deparaffinized, rehydrated, and incubated in PBS that contained 3% H₂O₂ for 15 min to block endogenous peroxidases. For TGF-β1 staining, tissue sections were incubated in PBS that contained 10% normal horse serum for 20 min to block nonspecific binding. Tissue samples were incubated with primary antibodies for 45 min. After incubation with a biotinylated IgG antibody, tissue sections were stained with a peroxidase-labeled streptavidin-biotin staining kit (DAKO GmbH, Hamburg, Germany) according to the manufacturer’s recommendations.

Primary antibodies included a tomato anti-lectin (Sigma-Aldrich Co., St. Louis, MO) as an endothelial marker, polyclonal rabbit anti–TGF-β1 (Santa Cruz Biotechnology, Santa Cruz, CA), polyclonal rabbit anti-CD3 (Novocastra Lab Ltd., Newcastle upon Tyne, UK) as a T cell marker, a polyclonal rabbit anti-VEGF (Santa Cruz Biotechnology), a monoclonal mouse anti-methylglyoxal (MGO), and a monoclonal mouse anti-carboxymethyllysine (CML; Biologo, Kronshagen, Germany). Replacement of the primary antibodies with PBS served as control.

Quantification of Histologic and Immunohistochemical Findings

For each visceral peritoneal specimen, >30 cross-sections were evaluated in a blinded manner. Interstitial fibrosis was analyzed using Sirius red–stained sections and a semiquantitative score system: 0, no interstitial fibrosis; 1, mild interstitial fibrosis; 2, moderate interstitial fibrosis with thickening of the peritoneum; and 3, severe interstitial fibrosis with marked thickening of the peritoneum. For determination of inflammatory cell number per area, CD3⁺ T cells per area and vessel number per area, all cells/vessels on a 121-point grid (Leitz, Wetzlar, Germany) were counted (area 0.028 μm²). For quantitative analysis of lectin, VEGF, TGF-β1, MGO, and CML, a 121-point grid (Leitz) was used. Points on stained visceral peritoneal tissue and all points on visceral peritoneal tissue were counted. The ratio in percentage was used to determine the degree of staining.

In Vitro Generation of AGE-Specific Fluorescence in PDF

The presence of AGE in unused PDF was assessed indirectly by measuring the generation of AGE-specific fluorescence after adding purified human serum albumin with a final concentration of 40 mg/ml (fraction V; Sigma Co., Taufkirchen, Germany). The samples were incubated without any preservatives at 37°C for 3 and 10 d. AGE concentration was assessed in unused and spent PDF recovered after 2 h as fluorescence intensity (excitation 350 nm/emission 430 nm) with a spectrofluorometer (LS 50 B; Perkin Elmer Co., Überlingen, Germany). At the end of experiments, sham i.p. animals received an injection of normal saline (pH 7.4). The analysis of measuring AGE fluorescence has been described in detail elsewhere (²⁰).

Electrophoretic Mobility-Shift Assay

Visceral peritoneal specimens were quick-frozen in liquid nitrogen, homogenized, transferred into cold buffer A (10 mM Hepes-KOH [pH 7.9, at 4°C], 10 mM KCl, 1.5 mM MgCl₂ 0.5 mM DTT, 1 mM EDTA, 0.2 mM PMSF, and 0.6% Nonidet P-40), incubated on ice for 10 min, and centrifuged for 5 min at 8000 rpm at 4°C. The supernatant was discarded, and the nuclear pellet was resuspended in 100 μl of buffer B (25% glycerol, 20 mM Hepes-KOH [pH 7.9 at 4°C], 420 mM NaCl, 1.5 mM MgCl₂, 0.2 mM EDTA, 0.5 mM dithiothreitol [DTT], 0.2 mM PMSF, 2 mM benzamidine, and 5 mg/ml leupeptin) and incubated on ice for 20 min. Cellular debris was removed by 2 min of centrifugation at 4°C, and the supernatant was quick-frozen at −80°C. Protein concentrations were determined by the Bradford assay. Nuclear extract (10 μg) was included in the binding reaction. Binding to an NF-κB consensus oligonucleotide (5′-AGTTGAGGGGACTTTCCAGGC-3′) was performed in 10 mM Hepes (pH 7.5) that contained 0.5 mM EDTA, 100 mM KCl, 2 mM DTT, 2% glycerol, 4% Ficoll 400, 0.25% Nonidet P40, 1 mg/ml BSA (DNAse free), and 0.3 μg/μl poly(dI/dC) in a total of 20 μl. Specificity of binding was ascertained by competition with a 160-fold molar excess of unlabeled consensus oligonucleotides and by characterization with specific polyclonal antibodies (Santa Cruz Biotechnology).

Statistical Analyses

All values are expressed as mean ± SEM. Wilcoxon and Kruskal-Wallis tests were used as appropriate to test statistical significance. Significance level was set at P < 0.05. Statistical analysis was performed by PC-Statistik (version 5.0; Hoffmann, Giessen, Germany) and GraphPad Prism (version 3.0.3; San Diego, CA).

Results

Animal Data

There were no significant differences with respect to age, body weight, and hematocrit values between RAGE −/− mice and WT mice (Table 1).

AGE Formation

In vitro, a time-dependent increase of AGE formation was demonstrated in low and high GDP PDF (P < 0.05 versus normal saline; Figure 1a) but not in filter-sterilized and glucose-free PDF. In vivo, however, AGE formation was not different in the various spent PDF (Figure 1b).

WT and RAGE −/− mice that were treated with either high or low GDP had an enhanced expression of MGO in peritoneal stainings compared with sham i.p. (P < 0.01; Figure 2a). CML was formed at an accelerated rate in peritoneal stainings of WT mice and RAGE −/− mice that were treated with high GDP but not in mice that were treated with low GDP (P < 0.01; Figure 2b).

Inflammation

In WT mice that were treated with high GDP, there was a significant increase in inflammatory cell number in visceral peritoneal tissue samples (P < 0.05 versus sham i.p.). This increase was not seen in WT mice that were treated with low GDP, filter-sterilized, and glucose-free PDF. RAGE −/− mice had a higher absolute inflammatory cell number in visceral peritoneal tissue samples at baseline; however, no significant increase was noted after treatment with GDP-containing PDF (Table 3).

These findings were confirmed by stainings for CD3⁺ T cells and Electrophoretic mobility-shift assay for NF-κB in selected groups (sham i.p., low GDP, and high GDP). There was a significant increase in CD3⁺ T cells and NF-κB binding activity in visceral peritoneal tissue samples of WT mice that were treated with high GDP (P < 0.05 versus WT sham i.p.; Figures 3 and 4). At baseline, a mild but statistically NS increase of NF-κB binding activity in peritoneal tissue of RAGE −/− mice was observed (Figure 4). After treatment with high GDP, no further increase could be detected. Furthermore, no increase of CD3⁺ T cell count was seen in RAGE −/− mice (Figure 3). Consistent with the above results, use of low GDP did not increase CD3⁺ T cell count and NF-κB binding activity in peritoneum of WT and RAGE −/− mice. Puncture trauma per se led to a significant higher inflammatory cell number when sham WT mice were compared with WT sham i.p. treatment (P < 0.05; Table 3).

Neoangiogenesis

WT mice that were treated with high GDP had an increased number of vessels and an enhanced expression of lectin and VEGF compared with sham i.p. (Table 3 and Figure 5, a and b). WT mice that were treated with low GDP had less pronounced neoangiogenesis than WT mice that were treated with high GDP (Figure 5, a and b). RAGE −/− mice showed a higher number of vessels per area at baseline but no further increase in number of vessels and in lectin and VEGF staining after long-term PDF administration (either filter-sterilized, glucose-free, low GDP or high GDP PDF; Table 3 and Figure 5, a and b).

Fibrosis

The fibrosis score was increased only in WT mice that were treated with high GDP but not in WT mice that were treated with low GDP and sham i.p. (P < 0.05; Table 3). Sham- and sham i.p.–treated WT mice were not significantly different. Using the Sirius red stain, an increase in fibrous tissue was found in the interstitium as well as in the vascular endothelium of WT mice that were treated with high GDP but not in RAGE −/− mice (Figure 6). Expression of TGF-β1 was more pronounced in WT mice that were treated with both high GDP and low GDP but not in sham i.p. (Figure 5c).

In contrast, RAGE −/− mice had no increase in fibrosis score at baseline and no further increase after long-term PDF treatment (either filter-sterilized, glucose-free, low GDP or high GDP PDF; Table 3). RAGE −/− mice showed higher TGF-β1 expression at baseline but no further increase after 12 wk of treatment with either high or low GDP (Figure 5c).

Discussion

Nondiabetic and nonuremic RAGE −/− mice did not show a progress of inflammation, neoangiogenesis, and fibrosis after prolonged twice-daily instillation of GDP-containing PDF. In contrast, in RAGE-bearing WT mice, a significant increase of inflammation (as indicated by CD3⁺ T cells and increased NF-κB binding activity), neoangiogenesis (accompanied by enhanced expression of lectin and VEGF), and fibrosis (increased expression of TGF-β1) was noted. This observation was more apparent in high as compared with low GDP, pointing to a dose-response relationship (^2,4–8).

Originally, high glucose concentrations and AGE were believed to be the main contributing factors in peritoneal damaging that result in PD failure (^1,10). In the meantime, toxic effects of carbonyl compounds (GDP) in PDF have been demonstrated (²¹). However, GDP lead to formation of AGE and enter the systemic circulation from the peritoneal cavity (²⁰). In addition, AGE have been detected immunohistochemically in the peritoneum of PD patients and likely are responsible for structural and functional peritoneal damage (^1,22). One hypothesis postulates that AGE-mediated peritoneal damaging seems to be dependent on the interaction with RAGE (¹⁸). RAGE is expressed in a variety of cell types, including endothelial cells, vascular smooth muscle cells, mesangial cells, and neuronal cells (^{13,14,16,23–25}). One study localized RAGE in the peritoneal membrane in vivo (¹⁷) and confirmed earlier observations of mRNA expression of functional RAGE in human peritoneal mesothelial cells in vitro (²⁶). In a diabetic animal model, inhibition of AGE–RAGE interaction with neutralizing monoclonal anti-RAGE antibodies prevents hyperglycemia-induced fibrosis of the peritoneal membrane (¹⁸). Although there is no doubt that RAGE is effectively blocked with anti-RAGE antibodies, it cannot be excluded presently that RAGE antibodies also affect structures that are different from RAGE. Other groups had demonstrated that application of sRAGE, an extracellular, truncated, soluble form of RAGE, in vivo and in vitro prevented late complications, i.e., microvascular and macrovascular complications in diabetes models. This raises the question of whether AGE-induced peritoneal damage can also be prevented by sRAGE. Remarkably, in each of these models (diabetic nephropathy, neuropathy, and arterial restenosis), protection from development of pathology was more profound in WT mice that were treated with sRAGE than in RAGE−/− mice (^27–29). In diabetic neuropathy, for example, administration of sRAGE to diabetic WT animals completely restored pain perception, whereas diabetic RAGE−/− mice were only partly protected from loss of pain perception (²⁸). These observations suggest that ligands that are sequestered by sRAGE are likely to interact with cellular structures other than RAGE. To avoid these methodologic culprits, we used a RAGE −/− mouse model to study the effects of high GDP PDF and low GDP PDF on the peritoneal membrane. We deliberately used nondiabetic and nonuremic mice to exclude enhanced AGE production and stimulation of RAGE by the uremic or the diabetic milieu.

AGE Formation

From our study we cannot entirely exclude a direct interaction of GDP products and RAGE, because RAGE is a known pattern recognition receptor that interacts with a variety of ligands. However, the key pathophysiologic step seems to be GDP-dependent AGE formation as shown by in vitro experiments (Figure 1a). Although measurement of AGE formation was not different in spent PDF (Figure 1b; most probably because of the short dwell time), there was an increased formation of CML in visceral peritoneal tissue samples of animals that were treated with high GDP (Figure 2b). These data indicate that rapid AGE modifications occur primarily on tissue-bound proteins.

For measurement of AGE formation, we used spectrofluorometry (^11,30). Although ELISA could be more sensitive, at least for an individual AGE compound, we prefer, in the sense of chemical identity, the more broad method to detect the sum of a variety of fluorescence AGE as discussed earlier (³¹).

Inflammation

RAGE was identified as a central signal transcription receptor for NF-κB activation (^32,33). In monocytes and in inflammatory lesions in various diseases (e.g., rheumatoid arthritis), increased RAGE expression was observed (^25,34). Members of the NF-κB family control the initiation of inflammation by regulating expression of cytokines and also play a central role in terminating inflammation (³⁵). The slight proinflammatory phenotype (also indicated by a slight increase in basal NF-κB binding activity and elevated cytokine levels) supports the hypothesis that RAGE might also be involved in the regulation of anti-inflammatory and antiproliferative processes (³⁶).

It is of note that the otherwise healthy RAGE −/− mice presented a higher baseline of CD3⁺ T cells, which corresponds with earlier reports describing a slight proinflammatory phenotype in the absence of RAGE (³⁶). This was also demonstrated in the peritoneum of RAGE −/− mice, in which NF-κB binding activity was slightly elevated when compared with WT mice. Despite a minor increased proinflammatory status, NF-κB binding activity did not further increase in RAGE −/− mice (low and high GDP), whereas NF-κB binding activity augmented significantly in the peritoneum of WT mice (Figure 4). In addition, in both RAGE −/− groups that were treated with high and low GDP, no further increase in CD3⁺ T cells was observed. The difference in the basal proinflammatory phenotype excludes the direct comparison of WT sham i.p. mice versus RAGE −/− mice that were treated with low and high GDP PDF. The overall observation that the increase in all parameters (inflammation, neoangiogenesis, and fibrosis) determined between sham i.p.–and GDP-treated animals is significant in WT mice but not in RAGE −/− mice clearly proves the involvement of RAGE. Recently, it was shown that RAGE −/− mice can mount a normal adaptive immune response. This was demonstrated in experimental autoimmune encephalomyelitis as an autoimmune model in which RAGE −/− mice developed the same inflammatory response as WT controls (¹⁹), therefore demonstrating that AGE–RAGE interaction must be the main contributing factor in transmitting GDP-dependent peritoneal inflammation. In addition, it should be considered that GDP-mediated toxicity is not only a local peritoneal phenomenon, because GDP enter the systemic circulation from the peritoneal cavity (²⁰). Both findings argue strongly against the use of PDF with a high GDP content. Compared with low GDP, filter-sterilized and glucose-free PDF showed no further benefit in reducing CD3⁺ T cells.

Neoangiogenesis

WT mice that were treated with either low or conventional high GDP presented an increased number of vessels per peritoneal area, which was accompanied by enhanced expression of VEGF. In vitro and in vivo exposure of peritoneal cells to high glucose stimulates the expression of VEGF (^34,37,38). In our study, exposure of the peritoneum to high GDP resulted in an increased staining for VEGF and increased vascular density. RAGE −/− mice demonstrated no increase of VEGF or vessels per peritoneal area with either low or high GDP. Although a correlation between AGE accumulation and VEGF expression had been demonstrated by de Vriese et al. (¹⁸), inhibition of AGE–RAGE interaction with anti-RAGE antibodies did not prevent peritoneal neoangiogenesis. In contrast to these observations, our results document a central role for GDP-induced upregulation of peritoneal neoangiogenesis by VEGF via AGE–RAGE interaction (^{12,15,24,32,39–42}) in the peritoneal membrane. As mentioned above, the absence of the proposed regulatory function of RAGE might also account for an increase of peritoneal vessels per area at baseline in RAGE-deficient mice.

Fibrosis

Exposure to high glucose concentrations induces collagen, fibronectin, or TGF-β1 production in various cells (^18,34,37). The role of high GDP concentrations in “extracellular” fibrosis in the peritoneum is less well understood. After 12 wk of treatment, no peritoneal fibrosis or upregulation of TGF-β1 expression was found in RAGE −/− mice. In contrast, fibrosis score and TGF-β1 expression were increased in WT mice that were treated with high GDP. Our data not only confirm the recent observations of de Vriese et al. that AGE–RAGE interaction plays an important role in the induction of extracellular fibrosis in PD in a diabetic mouse model (¹⁸) but also extend the hypothesis that AGE–RAGE interaction plays a pivotal role in the induction of peritoneal fibrosis in a nondiabetic (and nonuremic) mouse model.

In conclusion, the study presented here provides evidence that the RAGE is causally involved in the generation of structural and functional damage to the peritoneal membrane induced by GDP-containing PDF. These changes were less pronounced in low GDP PDF, indicating a dose-response relationship.

Figure 1: (a) Generation of fluorescence advanced glycation end-products (AGE) during in vitro incubation with albumin. Fluorescence intensity (excitation 350 nm/emission 430 nm) was quantified using a spectrofluorometer. *P < 0.05 versus normal saline on the same day; **P < 0.01 versus normal saline on the same day. (b) AGE fluorescence quantified with a spectrofluorometer (excitation 350 nm/emission 430 nm) in various spent peritoneal dialysis fluids (PDF).

Figure 2: Methylglyoxal (MGO; a) and carboxymethyllysine (CML; b) stainings of visceral peritoneal tissue in wild-type (WT) and receptor for AGE (RAGE)-deficient (RAGE −/−) mice after a 12-wk treatment period with either low or high glucose degradation products (GDP). Mice that were sham-injected without instilling solution as a control for puncture trauma (sham i.p.) received intraperitoneal injections twice daily. **P < 0.01 versus sham i.p. in the same group.

Figure 3: CD3⁺ stainings in WT and RAGE −/− mice after a 12-wk treatment period with either low or high GDP. Sham i.p. mice received intraperitoneal injections twice daily. *P < 0.05 versus sham i.p. in the same group.

Figure 4: Electrophoretic mobility-shift assay for NF-κB in visceral peritoneal tissue of WT and RAGE −/− mice after a 12-wk treatment period with either low or high GDP. Sham i.p. mice received intraperitoneal injections twice daily. *P < 0.05 versus sham i.p. in the same group.

Figure 5: Lectin (a), vascular endothelial growth factor (VEGF; b), and TGF-β1 (c) stainings in WT and RAGE −/− mice after a 12-wk treatment period with either low or high GDP. Sham i.p. mice received intraperitoneal injections twice daily. **P < 0.01 versus sham i.p. in the same group.

Figure 6: Sirius red staining of visceral peritoneal tissue from WT and RAGE −/− mice. (a) WT sham i.p. (b) WT mice after 12 wk of treatment with high GDP. (c) RAGE −/− sham i.p. (d) RAGE −/− after 12 wk of treatment with high GDP.

Table 1: Treatment groups and characteristics of animalsa

Table 2: Composition of different dialysis fluids

Table 3: Inflammation, neoangiogenesis, and fibrosis in WT and RAGE −/− mice

This work was supported in part by a grant from the Deutsche Forschungsgemeinschaft (DFG/SFB 405 to P.P.N.).

We thank Monika Klewer and Miriam Ramming (Department of Pathology, University of Erlangen-Nurnberg, Erlangen, Germany), Heike Ziebart (Department of Pathology, University of Heidelberg, Heidelberg, Germany), Ulrike Haug (Gambro Corporation Research, Hechingen, Germany), and Dr. Anders Wieslander (Gambro Corporate Research, Lund, Sweden) for excellent technical assistance.