Thrombospondin-1 Is an Endogenous Activator of TGF-β in Experimental Diabetic Nephropathy In Vivo Free

The animal studies were performed in accordance with the internal animal review board (Regierung von Mittelfranken: 621-2531.31-17/05). The animals were housed in a room maintained at 22 ± 2°C and exposed to a 12-h dark/12-h light cycle and fed standard mouse chow (Altromin 1324; Spezialfutterwerke, Lage, Germany) and tap water ad libitum. This study was performed in two different background strains comparing TSP-1–deficient with wild-type mice as previously described (19) to robustly test our hypothesis of TSP-1 as an important mediator of TGF-β activation during experimental diabetic nephropathy. In the first part of the study (7 and 9.5 weeks of diabetes), mice of C57Bl6 background (eight generations backcrossed) were used, and in the second part (20 weeks of diabetes), mice of 129SVJ background were used.

A scheme of the experimental protocol, including group sizes, is shown in Fig. 1. Experimental type 1 diabetes was induced by a single intraperitoneal injection of 180 mg STZ/kg body wt (Sigma, St. Louis, MO) injected in 50 mmol/l sodium citrate buffer (pH 5.5) using 3-month-old mice. Successful diabetes induction was first tested 1 week after induction followed by monthly blood glucose measurements. Nondiabetic control animals were treated with solvent alone. At least seven mice per group with repetitively confirmed high glucose levels (>400 mg/ml) were killed 7, 9.5, and 20 weeks after induction of diabetes. During the long-term (20 weeks) diabetes experiment, we transiently administered long-acting insulin (Lantus, human insulin HI-901; Sanofi Aventis, Frankfurt, Germany) subcutaneously at a dose of 2 units whenever the condition of animals was impaired or weight loss occurred. Using this procedure as described before (20–22), we successfully rendered animals hyperglycemic without becoming ketoacidotic. The day before the mice were killed, a 24-h urine collection was performed and proteinuria, albuminuria, and creatinine clearance were determined. On the day the mice were killed, blood was collected via vena cava puncture, kidneys were perfused with 0.9% sodium chloride followed by perfusion with Deltadex 40 including 0.05% Novocain, and the right kidney was harvested for immunhistochemical analysis. Afterward, the animals underwent perfusion fixation with 48 mmol/l Na₂HPO₄ and 14 mmol/l KH₂PO₄ including 3% glutaraldehyde for preparation of semi-thin sections (only done for the 20-week diabetes time point).

Tissue for light microscopy was fixed in methyl Carnoy’s solution or 4% paraformaldehyde, embedded in paraffin, and cut into 3-μm sections for indirect immunoperoxidase staining, as described previously (23). Sections were also stained with the periodic acid Schiff (PAS) reagent for determination of glomerular matrix expansion. Snap-frozen tissue was cut into 5-μm sections for detection of CD4⁺ and CD8a⁺ cells. Glomerular cell number was determined by counting hematoxylin-stained nuclei per 50 glomerular cross-sections within a biopsy. Glomerular hypertrophy was determined by measuring the glomerular tuft area of 60 glomerular cross-sections with computer-assisted morphometry using Metavue software (Visitron, Puchheim, Germany).

To perform immunostaining, tissue sections were incubated with the following primary and secondary antibodies: 19A2, a murine IgG monoclonal antibody (mAb) against the proliferating cell nuclear antigen (PCNA; Chemicon, Temecula, CA) (24), indicating actively proliferating cells; a rat monoclonal IgG2a to mouse CD4 antigen (Caltag Laboratories, Burlingame, CA,); a rat monoclonal IgG2a to mouse CD8a antigen (Caltag Laboratories) (24); F4/80, a murine IgG1 mAb to a surface receptor present on monocytes, macrophages, and dendritic cells (Caltag Laboratories); a polyclonal antibody to human collagen IV (Southern Biotechnology Associates, Birmingham, AL); fibronectin, a rabbit polyclonal antibody to rat fibronectin (Chemicon) (18); a murine IgG₁ mAb against TSP-1 (Dunn, Asbach, Germany) (25); a murine mAb for human desmin (DAKO, Glostrup, Denmark) (26); a rabbit polyclonal antibody to phosphorylated Smad2/3 (Santa Cruz Biotechnology, Santa Cruz, CA) (17); a chicken polyclonal anti-human active TGF-β1 (R&D Systems, Wiesbaden, Germany) (18); and a rabbit polyclonal to human plasminogen activator inhibitor-1 (PAI-1; Santa Cruz Biotechnology). Negative controls for immunostaining included either the omission of the primary antibody or substitution of the primary antibody with equivalent concentrations of an irrelevant murine mAb or preimmune rabbit IgG. All immunoperoxidase stainings was done using methyl Carnoy’s fixed tissues. After incubation with primary antibodies overnight at 4°C, specific biotinylated secondary antibodies (all by Vector Laboratories, Burlingame, CA) were applied to tissue sections, followed by peroxidase-conjugated Avidin D (Vector Laboratories) and color development with diaminobenzidine, with or without nickel chloride, for nuclear staining.

For frozen sections, a Cy3-labeled donkey anti-rat IgG (Jackson ImmunoResearch Laboratories, West Grove, PA) was used as secondary antibody. For each analysis, 25–50 glomerular cross-sections were evaluated in a blinded fashion. Glomerular active TGF-β and expression of fibronectin was quantified by computerized measurement of the positively stained glomerular area using the Metavue Imaging System (Visitron). Mesangial matrix expansion was graded on a scale of 0 to 4 (0, normal/no staining; 1, <25%; 2, 26–50%; 3, 51–75%; and 4, >75% of the glomerulus involved or evaluated in semi-thin sections using a 10 × 10 grid overlaying each glomerulus and counting the number of squares completely filled out by PAS⁺ mesangium). Collagen IV was graded semiquantitatively and reflected changes in the area and intensity of mesangial staining: 0, very weak or absent staining; 1+, weak staining with <25% of the glomerular tuft showing focally increased staining; 2+, 25–49% of the glomerular tuft with focally increased staining; 3+, 50–75% of the glomerular tuft demonstrating increased staining; and 4+, >75% of the glomerular tuft stained strongly. In addition, glomerular PAI-1 expression was also graded semiquantitatively on a scale of 0 to 4 (0, absent staining; 1, weak podocytic staining; 2, strong podocytic staining; 3, podocytic and mesangial staining; and 4, podocytic and mesangial staining of >50% of the glomerular tuft). For slides immunostained for desmin, only the podocytes at the outer edge of the glomerular tuft were assessed as described elsewhere (26). F4/80⁺, CD4⁺, and CD8a⁺ cells were analyzed in 20 fields (magnification ×400).

Apoptotic cells were detected by the terminal deoxynucleotidyl transferase (TdT)-mediated dUTP-biotin nick-end labeling (TUNEL) assay, as described previously (27,28).

RNA was isolated from laser microdissected glomeruli using snap-frozen biopsies cut into 10-μm sections, extracted by RLT buffer, and purified using RNeasy columns (both from Qiagen, Hilden, Germany). For each sample, at least 100 laser microdissected glomeruli were used. Subsequently, RNA was treated with DNase to avoid DNA contamination. Reverse transcription was performed using TaqMan reverse transcription reagents (Applied Biosystems, Weiterstadt, Germany) following manufacturer’s instructions.

Real-time RT-PCR was performed on a TaqMan ABI 7000 sequence detection system using the Mastermix (all Applied Biosystems). After an initial hold of 3 min at 95°C, samples were cycled 40 times at 95°C for 15 s and 60°C for 60 s. The cDNA content of each sample was compared with 18S (forward primer, 5′- TTGATTAAGTCCCTGCCCTTTGT-3′; reverse primer, 5′- CGATCCGAGGGCCTCACTA-3′) (29) as a housekeeping gene following the ΔΔCt technique. For real-time PCR, the following primers were used: fibronectin (forward primer, 5′-TGTGACCAGCAACACGGTG-3′; reverse primer, 5′-ACAACAGGAGAGTAGGGCGC-3′); TGF-β1 (forward primer, 5′-TGACGTCACTGGAGTTGTACGG-3′; reverse primer, 5′-GGTTCATGTCATGGATGGTGC-3′; probe, 5′-TTCAGCGCTCACTGCTCTTGTGACAG-3′) (30); and TGF-β2 (QuantiTech Primer Assay; Qiagen).

Urinary protein was measured using the Bio-Rad Protein Assay (Bio-Rad, München, Germany) with BSA (Sigma) as a standard. Creatinine in serum or urine and blood urea nitrogen were measured using an autoanalyzer (Beckman, Munich, Germany). Albuminuria was determined using an enzyme-linked immunosorbent assay kit (Axxora, Grünberg, Germany) following the manufacturer’s instructions.

All values are expressed as means ± SD. Statistical significance (defined as P < 0.05) was evaluated using the Student’s t test or one-way ANOVA with modified t test using the Bonferroni method.

TSP-1 is expressed in diabetic glomeruli of wild-type mice showing a typical mesangial expression pattern 20 weeks after induction of diabetes (Fig. 2A) as demonstrated by immunohistological staining (Fig. 2C). No staining was observed in glomeruli from TSP-1–deficient mice (Fig. 2B) or healthy animals (data not shown). Average blood glucose levels in healthy control animals were detected at ∼100 mg/dl and increased to 400–500 mg/dl after induction of diabetes persisting during the whole experiment (Fig. 2D). Renal hypertrophy could be seen in all diabetic animals, as demonstrated by an increase of the kidney weight–to–body weight ratio of >40% (Fig. 2E). Neither blood glucose levels nor renal hypertrophy was influenced by the lack of TSP-1. Glomerular enlargement during diabetes, as measured by the average glomerular area from cross-sections, was lower in TSP-1–deficient mice compared with wild-type animals, reaching significance after 9.5 weeks but not after 20 weeks (P = 0.24) (Fig. 2F).

If the de novo expressed TSP-1 is activating TGF-β in STZ-induced diabetic nephropathy, glomerular TGF-β activity but not total TGF-β expression should be reduced in TSP-1–deficient mice. Active TGF-β in glomeruli injured by diabetes was determined by using several different readouts. First, immunostaining using an antibody specifically recognizing the active form of TGF-β1 was performed (Fig. 3A and B). Second, immunostaining was done with an antibody specific for the phosphorylated form of the TGF-β signal transduction molecules Smad2/3 (Fig. 3C and D). Finally, staining was done for the TGF-β target gene PAI-1 (Fig. 3E and F). The amount of active TGF-β steadily increased with duration of diabetes in wild-type mice reaching highest active TGF-β amounts on week 20 (Fig. 3A). TSP-1 deficiency was associated with inhibition of increased glomerular TGF-β activity 9.5 and 20 weeks after induction of diabetes (Fig. 3A; P < 0.0005). Consequently, the increased percentage of glomerular cells positive for the phosphorylated form of the TGF-β signaling molecule Smad2/3 was substantially lower in TSP-1–deficient diabetic mice at all time points investigated (Fig. 3D; P < 0.0005). TSP-1–deficient mice also showed significantly lower glomerular PAI-1 expression at all time points after diabetes induction (Fig. 3C; P < 0.02). In contrast, total glomerular TGF-β was similar in wild-type and TSP-1–deficient diabetic mice, as assessed by immunostaining for TGF-β1 and -2 on cross sections (Fig. 4A and B) and by real-time PCR for TGF-β1 and -2 mRNA from microdissected glomeruli 20 weeks after diabetes induction (Fig. 4C and D). Western blot analysis for TGF-β1 and TGF-β2 using cortical extracts from mice after 9.5 weeks of diabetes (data not shown) also showed increased levels associated with diabetic nephropathy but no differences in wild-type versus TSP-1–deficient mice.

Reduction in TGF-β activity was accompanied by reduced glomerular matrix accumulation in diabetic TSP-1 knockout (KO) mice. After 7 weeks of diabetes, glomerular matrix accumulation started, which was much more prominent at week 20 (Fig. 5A). Glomerular PAS positivity was significantly lower 7 and 20 weeks after diabetes induction in mice lacking TSP-1 (Fig. 5A). Because matrix accumulation was most prominent in mice that were diabetic for 20 weeks, we additionally used semi-thin sections for evaluation of mesangial matrix expansion at this point (Fig. 5B). In glomeruli from wild-type mice, mesangial matrix expansion was typically quite prominent 20 weeks after diabetes induction (Fig. 5B and C). In contrast, the glomeruli from diabetic TSP-1–deficient mice showed minor glomerular changes compared with healthy controls (Fig. 5D and E) demonstrating a significant reduction of matrix accumulation versus wild-type mice after 20 weeks of diabetes (Fig. 5B; P < 0.02). Consistent with this result, glomerular fibronectin staining was significantly reduced in TSP-1–deficient diabetic mice at each time point investigated (Fig. 5F; P < 0.003), and fibronectin mRNA was almost significantly reduced 20 weeks after diabetes induction (Fig. 5G; P = 0.06). Constitutive expression of glomerular collagen IV in healthy animals only slightly but significantly (P < 0.03) increased after diabetes induction in wild-type animals. This increased collagen IV expression during diabetes was prevented in TSP-1–deficient mice 7 and 9.5 weeks after diabetes induction and significantly reduced compared with wild-type animals on week 20 (Fig. 5H).

Not only mesangial cells but also podocytes are an important target of renal diabetic injury. Because desmin has been established as an excellent marker of podocyte injury during diabetic nephropathy (26), podocytic desmin expression was evaluated in wild-type and TSP-1–deficient diabetic mice. Podocytic desmin expression was significantly lower in TSP-1–deficient mice 20 weeks after diabetes induction, indicating less podocyte damage in these animals (Fig. 5I; P < 0.05).

Inflammatory cells are important mediators for the induction of renal fibrosis. Therefore, influx of macrophages and T-cells were investigated in wild-type and TSP-1–deficient mice during STZ-induced diabetic nephropathy. All inflammatory cell types assessed (macrophages via F4/80 staining, T-cells via CD4 and CD8 staining) were increased within the renal cortex during diabetic nephropathy. Accumulation of these different inflammatory cell types within the renal cortex was overall lower in TSP-1–deficient mice, reaching statistical significance at 9.5 weeks (C57Bl6 background) and variably at 20 weeks (129 SVJ background) after diabetes induction. Accumulation of cortical macrophages was significantly lower in diabetic TSP-1–deficient mice at 9.5 but not 20 weeks after diabetes induction because of a high SD (Fig. 6A). Cortical influx of T-cells as assessed by CD4 and CD8a positivity was significantly reduced by 52 and 64% in TSP-1–deficient mice 9.5 weeks after diabetes induction compared with wild-type controls (Fig. 6B and C). In parallel to cortical macrophages, T-cell infiltration was more than four times higher compared with healthy controls in wild-type mice after 20 weeks of diabetes. At that time point, significantly fewer CD4⁺ cells were again found in TSP-1–deficient mice, but CD8a⁺ cells only showed a tendency to lower numbers (Fig. 6B and C) compared with wild-type diabetic mice.

Because TSP-1 is known to influence both cellular proliferation and apoptosis, these measures were investigated during diabetic nephropathy. During the first 10 weeks of diabetes, glomerular cell number increased only slightly and to an equal extent in wild-type and TSP-1–deficient mice (C57Bl6 background). Twenty weeks after diabetes induction (129SVJ background), a significant increase of glomerular cell number was detected in wild-type mice, which was ameliorated in TSP-1–deficient mice (Fig. 7A; P < 0.0002). Differences in cell number are due to the balance of local proliferative and apoptotic activity and influx of systemic cell types. Glomerular proliferation was slightly enhanced during diabetes in both groups starting on different levels in the two different backgrounds. Independent of the genetic background, neither the proliferative (Fig. 7B) nor the apoptotic activity (as assessed by TUNEL assay; Fig. 7C and D) within glomeruli or cortical areas was significantly different in TSP-1–deficient versus wild-type mice.

Proteinuria, a marker of severity of diabetic nephropathy, was unchanged during early diabetes in both groups, indicating only a minor impairment of renal function at the beginning of diabetic nephropathy. Twenty weeks after diabetes induction, proteinuria in diabetic mice was increased by approximately eightfold compared with healthy controls. This response was decreased by ∼50% in TSP-1–deficient mice compared with wild-type animals (Fig. 8A; P < 0.003). Albuminuria, an even more sensitive marker of renal diabetic injury, was increased during early diabetes in the C57bl6 background (weeks 7–9.5) and was maximally increased 20 weeks after diabetes induction in the 129 SVJ background (Fig. 8B). TSP-1–deficient mice showed a tendency to lower albuminuria levels during early diabetes (7–9.5 weeks), which almost reached significance at week 20 (Fig. 8B; P = 0.08). Serum urea as another indicator of renal functional impairment was also increased during diabetic nephropathy. Serum urea was significantly reduced in TSP-1–deficient mice 9.5 weeks after diabetes induction and showed an almost significant tendency to lower levels 20 weeks after diabetes induction (Fig. 8C; P = 0.06). Creatinine clearance was the same in both groups (9.12 ± 2.8 in wild-type mice vs. 8.26 ± 3.4 ml/kg body wt in TSP-1–deficient mice), but only the colorimetric and not a more accurate high-performance liquid chromatography–based methodology was used.

Diabetic nephropathy already is the leading cause for end-stage renal disease in the Western world (1,2,4), and the incidence is increasing. Many in vitro and in vivo studies confirmed TGF-β as a key player in the pathogenesis of diabetic nephropathy (6).

This study, for the first time, shows that TSP-1 is an endogenous activator for TGF-β in diabetic nephropathy in vivo. Deficiency for TSP-1 in diabetic mice reduced activation but not expression of TGF-β in our study. Although glomerular levels of TGF-β1 and TGF-β2 protein and mRNA of diabetic wild-type and TSP-1 KO mice were equivalent, TGF-β activation was reduced in diabetic TSP-1 KO mice, as indicated by several different methods for assessing TGF-β activity directly (specific antibody) or indirectly, via the degree of phosphorylation of the TGF-β signaling molecule smad2/3, as well as the product of the transcriptional TGF-β target, PAI-1. Reduction of TSP-1–mediated TGF-β activation in diabetic nephropathy apparently is able to ameliorate several key features typical of the pathophysiology of diabetic nephropathy, such as increased matrix accumulation and glomerulosclerosis, podocyte injury, glomerular hypertrophy, and proteinuria.

The data presented here suggest that TSP-1 is an important activator of TGF-β in the STZ-induced diabetic nephropathy model. This interpretation is supported by the fact that some diabetic changes (matrix accumulation) in diabetic TSP-1 KO mice are close to healthy controls and the fact that the pattern and the degree of inhibition of diabetic nephropathy features are very similar to studies using general TGF-β–blocking strategies (antibodies) in short-term type 1 (31) or long-term type 2 experimental diabetic nephropathy (32).

Several recent studies demonstrate the importance of inflammation during diabetic nephropathy (33–35). Therefore, the finding of this study that influx of several different inflammatory cell types, such as macrophages and CD4⁺ and CD8a⁺ cells, is ameliorated in TSP-1–deficient diabetic mice, compared with wild-type mice suggests that a decreased inflammatory response is the cause of the protective effect of TSP-1 deficiency. It is unclear whether this effect is due to altered active TGF-β levels or to other TSP-1–dependent mechanisms. As discussed in the literature, the role of TGF-β on inflammation is controversial. TGF-β has anti-inflammatory properties as demonstrated by several experimental studies in different models (36,37) and by the fact that TGF-β1–deficient mice die because of multifocal inflammatory disease (38), whereas TGF-β overexpressing mice exert beneficial anti-inflammatory actions of TGF-β (39). On the other hand, TGF-β can induce macrophage chemoattractant protein 1 in mesangial cells (40). TSP-1 itself is known to promote chemotaxis of leukocytes to inflammatory sites and has a central role in activation and clonal expansion of inflammatory T-cells (41), which would be consistent with the results of our study.

Nevertheless, the limitations of this study need to be discussed. First, participation of other activators of TGF-β or direct secretion of the active cytokine by glomerular cells cannot be completely excluded. Second, after a single high dose of STZ, substantial collateral tissue toxicity may occur, which increases albuminuria (42). TSP-1–mediated effects on destruction/regeneration of β-cells are unlikely, because no differences in blood glucose levels were detected at each time point, but they cannot completely be excluded. Third, our study cannot differentiate whether effects on diabetic nephropathy are due to a systemic, bone marrow–derived, or local lack of TSP-1. Fourth, we show equivalent effects of TSP-1 deficiency in diabetic nephropathy in mice of two different backgrounds, but direct comparison of changes during the time course is invalid because different time points were investigated. Fifth, our STZ-mediated diabetic nephropathy model compared with late diabetic nephropathy in humans showed relatively mild lesions with predominant diabetic glomerulopathy, limiting the conclusions toward human disease. Nevertheless, the 20-week time point of STZ-induced diabetic nephropathy demonstrated clear-cut diabetic glomerular changes that are ameliorated by TSP-1 deficiency. Early treatment or even prophylaxis of diabetic nephropathy in humans would be ideal because TSP-1 expression correlates with disease severity during diabetic nephropathy (16). Furthermore, in humans, only 30% of all diabetic patients develop diabetic nephropathy, suggesting genetic progression factors. In a recent study, peripheral mononuclear blood cells from diabetic patients with and without diabetic nephropathy were screened/compared for candidate target genes using microarray technology. TSP-1 was one out of four differentially regulated genes, suggesting that TSP-1 expression is important for the pathogenesis of diabetic nephropathy in humans (43). In addition, the fact that we could show similar results regarding TSP-1–mediated TGF-β activation and matrix accumulation using mice of two different backgrounds strongly supports the relevance of these findings.

In conclusion, the studies described above identify TSP-1 as an endogenous activator of TGF-β in murine type 1 diabetic nephropathy. Prevention of TSP-1–mediated TGF-β activation in diabetic nephropathy is able to ameliorate several key features typical for the pathophysiology of diabetic nephropathy such as increased matrix accumulation and glomerulosclerosis, podocyte injury, glomerular hypertrophy, and proteinuria. Targeting TSP-1–mediated TGF-β activation is therefore a new therapeutic option to ameliorate diabetic nephropathy.

This study has received support in part from the Deutsche Forschungsgemeinschaft (SFB 423, TP B6) and from the Interdisciplinary Center for Clinical Research (Project A12).

The technical help of Ulrike Goller, Andrea Braun, Andrea Lüdke, Susanne Weber, and Tanja Christ is gratefully acknowledged.

Parts of this study were presented at the 38th annual meeting of the American Society of Nephrology, Philadelphia, 8–13 November 2005.