Podocyte-Derived Vascular Endothelial Growth Factor Mediates the Stimulation of α3(IV) Collagen Production by Transforming Growth Factor-β1 in Mouse Podocytes

Conditionally immortalized mouse podocytes were kindly provided by Dr. Peter Mundel (Albert Einstein College of Medicine, Bronx, NY) and were handled as previously described (21). The cells harbor a temperature-sensitive variant of the SV40 large T antigen (tsA58) that is inducible by γ-interferon and stable at 33°C but rapidly degraded at 37–39°C (31). At 33°C, the large T antigen allows for cellular proliferation. The growing cells are maintained in collagen I–coated flasks in RPMI-1640 media supplemented with 10% FCS, 100 units/ml penicillin, 100 μg/ml streptomycin, and 10 units/ml mouse recombinant γ-interferon. When the cells have reached confluence, they are passaged and allowed to differentiate at 37.5°C for 2 weeks without γ-interferon in Dulbecco’s modified Eagle’s medium containing 5.5 mmol/l glucose, 5% FCS, 100 units/ml penicillin, and 100 μg/ml streptomycin. Podocyte differentiation is evidenced by the expression of α-synaptopodin (31).

Differentiated podocytes were treated with exogenous recombinant mouse VEGF₁₆₄ (R&D Systems, Minneapolis, MN) in concentrations of 2, 8, and 20 ng/ml for 24 or 48 h. Recombinant human TGF-β1 (R&D Systems) was added at 2 ng/ml for 48 h. This dose and duration has been shown to significantly stimulate endogenous VEGF secretion and α3(IV) collagen production as well as inhibit α5(IV) collagen production in mouse podocytes (21). Concurrent with the VEGF or TGF-β1 treatment, the following specific inhibitors were added in certain experiments: a pan-VEGF receptor inhibitor (SU5416; gift of SUGEN/Pfizer; IC₅₀ [half-maximal inhibitory concentration] ∼1 μmol/l for inhibition of receptor autophosphorylation) (32), a PI3K inhibitor (LY294002; Calbiochem, San Diego, CA), or a TGF-β1 receptor kinase inhibitor (SB-431542; GlaxoSmithKline, King of Prussia, PA) (21, 33).

Preparation of total lysate protein in radioimmunoprecipitation assay buffer followed by SDS-PAGE (3–8 or 4–12% gradient; NuPAGE precast gels; Invitrogen, Carlsbad, CA) and wet transfer to a nitrocellulose membrane (Bio-Rad, Hercules, CA) were performed as previously described (21). The membrane was blocked in 5% nonfat milk in Tris-buffered saline/0.1% Tween 20 and probed overnight at 4°C with one of the following primary antibodies: human anti-α3(IV) collagen (gift of Dr. Michael Madaio, University of Pennsylvania), rabbit anti-α5(IV) collagen (gift of Dr. Jean-Francois Beaulieu, University of Sherbrooke, Québec, Canada), rabbit anti–VEGFR-1 (NeoMarkers/LabVision, Fremont, CA), rabbit anti–VEGFR-2 (NeoMarkers), rabbit anti–phospho-VEGFR-1 (Oncogene Research Products, San Diego, CA), rabbit anti-Smad2 (Zymed, South San Francisco, CA), rabbit anti–phospho-Smad2 (Upstate Biotechnology, Charlottesville, VA), or mouse anti–β-actin antibody (Sigma, St. Louis, MO). In some cases, the blocking peptide for VEGFR-1 and -2 (both NeoMarkers) or clusterin (Santa Cruz Biotechnology, Santa Cruz, CA) was incubated overnight with the corresponding antibody before use. After three washes in Tris-buffered saline/0.1% Tween 20, membranes were probed with a secondary goat anti-human (Jackson ImmunoResearch, West Grove, PA), donkey anti-rabbit (Amersham Biosciences, Piscataway, NJ), or sheep anti-mouse antibody (Amersham), each conjugated to horseradish peroxidase (HRP). The HRP-catalyzed chemiluminescence reaction was developed with SuperSignal West Pico substrate (Pierce Biotechnology, Rockford, IL), allowing the detection of immunoreactive protein bands. The membrane was reprobed with β-actin to correct for small differences in loading. The resulting bands on film were quantitated by computer-assisted video densitometry. Using the ImageJ 1.29x software (National Institutes of Health, Bethesda, MD), the protein band density was measured. The amount of protein under control conditions was assigned a relative value of 100%.

Cell culture media supernatants were frozen at −20°C until assayed for VEGF using a commercial enzyme-linked immunosorbent assay (ELISA) kit (R&D Systems). The supernatants were dispensed onto a microtiter plate coated with an anti-VEGF antibody that recognizes mouse VEGF_{120, 164} and then sandwiched with a secondary anti-VEGF antibody conjugated to HRP. The resulting chromogenic reaction was stopped and read at 450 nm, and the concentration of VEGF was determined from the standard curve. VEGF concentrations were then corrected for the total protein concentrations and the data expressed as picogram VEGF per milligram protein.

Total RNA was purified with TRIzol (Invitrogen) from both dividing undifferentiated and differentiated podocytes. As a positive control, total RNA was also extracted from the cortex of a mouse kidney. The RNA was reverse transcribed to cDNA with avian myeloblastosis virus, random hexamers, and deoxynucleotides, as directed in the SuperScript kit (Invitrogen). The PCR primers for VEGFR-1 or -2 were added, along with deoxynucleotides and Taq polymerase (Invitrogen). For mouse VEGFR-1, the forward primer was 5′-CTC TGA TGG TGA TCG TGG-3′ and the reverse primer 5′-CAT GCC TCT GGC CAC TTG-3′ (34). For mouse VEGFR-2, the forward primer was 5′-GCC AAT GAA GGG GAA CTG AAG AC-3′ and the reverse primer 5′-TCT GGC TGC TGG TGA TGC TGT C-3′ (modified from the rat sequence) (35). PCR was performed under the following conditions: denaturation at 94°C for 2 min, 40 cycles of denaturation at 94°C for 1 min, annealing at 58°C for VEGFR-1 and 53°C for VEGFR-2 for 30 s, and extension at 72°C for 1 min, followed by final extension at 72°C for 4 min. An aliquot of the final PCR product was run in a 2% agarose gel stained with ethidium bromide. The predicted band sizes were 317 bp for VEGFR-1 and 538 bp for VEGFR-2. The remainder of the PCR product was cleaned up with a PCR purification kit (QIAquick; Qiagen, Valencia, CA) and sequenced in both directions with the forward and reverse primers (University of Pennsylvania DNA Sequencing Facility). The elucidated sequence of the RT-PCR product was compared with the published mRNA sequence with the aid of a sequence analysis program (MacVector; Accelrys, San Diego, CA).

Podocytes were seeded onto collagen I–coated four-chamber slides (Nunclon, Roskilde, Denmark) and allowed to differentiate over 2 weeks at 37.5°C. Chambers were washed three times in 1× PBS for 5 min. Cells were then fixed in a 50:50 mixture of methanol:acetone for 8 min at −20°C. Afterward, 0.1% triton X-100 in PBS was added for 3 min to permeabilize the cells. Nonspecific binding sites were blocked with 1% BSA in PBS for 1 h at room temperature. The four chambers were treated with the following primary antibodies at a 1:500 dilution in 1% BSA: 1) anti–VEGFR-1 antibody alone, 2) anti–VEGFR-1 antibody preabsorbed overnight with 10 times the volume of VEGFR-1 blocking peptide, 3) anti–VEGFR-1 antibody preabsorbed with 10 times the volume of clusterin (irrelevant blocking peptide), and 4) no primary antibody. This setup was duplicated for the anti–VEGFR-2 antibody. After a 2-h incubation at room temperature, the chambers were washed three times in 1× PBS, and a Cy3-conjugated donkey anti-rabbit antibody (Jackson ImmunoResearch) was added at a 1:1,000 dilution in 1% BSA. The secondary antibody was allowed to incubate in the dark for 1 h at room temperature. After three more 5-min washes in 1× PBS and one brief wash in dH₂O, a 1:10,000 dilution of 4′,6-diamidino-2-phenylindole (DAPI) (Molecular Probes, Eugene, OR) was added to each chamber for 30 s to stain nuclei. The plastic chambers were then removed and the slide mounted under a coverslip in fluorescence mounting medium (Kirkegaard & Perry Laboratories, Gaithersburg, MD). Photomicrographs of each section of the slides were taken at 200× magnification with a fluorescence microscope (Nikon Eclipse E600) connected to a charge-coupled device digital camera (CoolSNAP cf.; Roper Scientific, Trenton, NJ). Fluorescence was imaged separately for standardized times in the red and blue channels for Cy3 and DAPI. A brightfield image was obtained using differential interference contrast (or Nomarski) microscopy. Finally, the fluorescent and brightfield images were electronically merged using the IPLab Mac software (Scanalytics, Fairfax, VA).

Graphical data are displayed as means ± SE for the number of independent experiments indicated in the figure legends. Unpaired Student’s t test was used to compare the control with the experimental group. P < 0.05 was considered statistically significant.

Cultured podocytes were exposed to exogenous mouse in doses of 2, 8, and 20 ng/ml VEGF₁₆₄. By Western blot analysis, all of the VEGF doses significantly stimulated the production of α3(IV) protein by 52–73% at 24 h (Fig. 1A). When exogenous VEGF stimulation was continued for 48 h, α3(IV) collagen production reached a peak (80% increase) at 8 ng/ml VEGF then tapered off at 20 ng/ml (Fig. 1A). The control level of α3(IV) collagen at 48 h was greater than at 24 h, suggesting the ongoing production of collagen over time. The degree of VEGF-stimulated α3(IV) production was not affected by SB-431542, a TGF-β signaling inhibitor (33) (n = 6, P = NS for VEGF vs. VEGF plus SB-431542).

Similarly designed studies were conducted to assay the effect of VEGF on the production of α5(IV) collagen measured by immunoblotting. At the VEGF doses and treatment times used, VEGF had no significant effect on the production of α5(IV) collagen (Fig. 1B).

The VEGF receptor tyrosine kinase proteins are specifically blocked by SU5416 at an IC₅₀ of ∼1 μmol/l for the inhibition of VEGFR-1 and -2 autophosphorylation (15, 32). SU5416 dose-dependently diminished the production of α3(IV) collagen in the basal state, reaching a plateau at ∼5 μmol/l SU5416 (Fig. 2A). Downstream of the receptors, VEGF can signal through the PI3K-Akt pathway, so the PI3K inhibitor, LY294002, was also tested; this treatment progressively decreased the basal production of α3(IV) protein (Fig. 2B), similar to SU5416 (Fig. 2A). When 8 ng/ml exogenous VEGF was added for 48 h, the increase in α3(IV) protein was completely prevented by the concurrent addition of 1 μmol/l SU5416 (Fig. 2C). This intermediate dose of SU5416 did not reduce the basal amount of α3(IV) collagen. However, a higher dose of SU5416 (5 μmol/l) not only prevented exogenous VEGF from stimulating α3(IV) production but also depressed the basal α3(IV) collagen level (Fig. 2C), consistent with the notion that SU5416 can inhibit the effect of endogenously produced VEGF.

The RT-PCR product for VEGFR-1 was detected at the appropriate size in differentiated mouse podocytes, but none were detected for VEGFR-2 (Fig. 3A). Failure to identify VEGFR-2 was not due to technical problems because the correct-sized band for VEGFR-2 was readily amplified from mouse kidney cortex (Fig. 3A). In addition, RT-PCR did not identify VEGFR-2 mRNA in dividing undifferentiated podocytes (data not shown). To prove that the VEGFR-1 band had originated from mRNA and not from contaminating genomic DNA, the addition of RNase before the RT-PCR procedure completely abrogated the band (data not shown). Finally, the VEGFR-1 band was sequenced and found to align perfectly with the published mRNA sequence for mouse VEGFR-1 (gi:34328179) (Fig. 3B).

The Western blot analysis for VEGFR-1 showed a band at ∼180 kDa (Fig. 4A). However, preincubation of the anti–VEGFR-1 antibody with increasing doses of a specific blocking peptide before immunoblotting progressively competed away the intensity of the band (Fig. 4A). Moreover, the addition of an irrelevant blocking peptide, clusterin, did not dampen the intensity of the band (Fig. 4A), proving the specificity of the antibody for VEGFR-1.

In contrast, the Western blot analysis for VEGFR-2 showed a band that was not at the expected size of 180 kDa but at 110 kDa (Fig. 4B), which was likely the result of nonspecific binding of the anti–VEGFR-2 antibody since the band did not fade with increasing doses of the blocking peptide (Fig. 4B). Thus, VEGFR-2 protein is probably absent from cultured mouse podocytes, which is consistent with the RT-PCR data (Fig. 3A).

Corroborative immunostaining for the VEGF receptor proteins was performed in the podocyte. The stain for VEGFR-1 protein showed a strong fluorescent signal (red color) localized to the cell body of the podocyte (Fig. 5A). Furthermore, preincubation of the anti–VEGFR-1 antibody with a blocking peptide largely abrogated the immunostaining (Fig. 5B), decreasing the fluorescent signal down to background levels seen with the negative control, in which the primary antibody was omitted (Fig. 5D). Finally, the addition of an irrelevant blocking peptide, clusterin, to the anti–VEGFR-1 antibody did not compete away the intensity of the immunofluorescence (Fig. 5C). In contrast, the immunostaining for VEGFR-2 was almost nil under all the specificity conditions tested as above (Fig. 5E–H).

Autophosphorylation of VEGFR-1 was sought as evidence for the initiation of VEGF signaling in podocytes. After stimulation with recombinant mouse VEGF₁₆₄ for 0.5 h, the amount of phospho–VEGFR-1 was detectable above background (Fig. 6). With continued VEGF stimulation, VEGFR-1 became increasingly phosphorylated, reaching a peak at ∼18 h (Fig. 6). Elevated phospho–VEGFR-1 levels were sustained for the remainder of the experiment. On the other hand, total VEGFR-1 and β-actin levels stayed relatively constant (Fig. 6).

Consistent with our previous data (21), addition of 2 ng/ml TGF-β1 for 48 h increased the production of endogenous VEGF by 2.7-fold; the media supernatant level of mouse VEGF_{120, 164}, measured by ELISA (R&D Systems), increased from 111 ± 8.5 pg VEGF/mg protein to 294 ± 31.4 pg VEGF/mg protein (n = 5, P < 0.001 vs. control). Concomitantly, 2 ng/ml TGF-β1 stimulated α3(IV) protein production by 2.3-fold (Fig. 7A), a response that was completely prevented by pretreatment with 1 μmol/l SB-431542, a direct inhibitor of TGF-β signaling (33). On the other hand, stimulation of α3(IV) collagen production by TGF-β1 was partially and significantly prevented by pretreatment with either 5 μmol/l SU5416 (VEGF receptor inhibitor) or 25 μmol/l LY294002 (PI3K inhibitor). Additionally, Fig. 7A shows that α3(IV) collagen levels under control conditions were decreased by treatment with SU5416 or LY294002, further suggesting that endogenous VEGF maintains a small component of the basal α3(IV) production in podocytes (Fig. 2A–C).

In contrast, 2 ng/ml TGF-β1 decreased α5(IV) collagen levels in mouse podocytes (Fig. 7B) as previously described (21). TGF-β1–mediated suppression of α5(IV) collagen was not significantly changed by SU5416 treatment (Fig. 7B), which is consistent with the data that VEGF had no effect on the production of α5(IV) collagen (Fig. 1B). However, the inhibitory effect of TGF-β1 on α5(IV) production could be successfully reversed by the TGF-β signaling inhibitor, SB-431542 (Fig. 7B), indicating that the effect of TGF-β1 on α5(IV) is dependent upon the TGF-β receptor system but not the VEGF receptor system.

To exclude the possibility that either SU5416 or LY294002 was interfering directly with the TGF-β1 signaling system, we assessed the effects of these inhibitors on the phosphorylation of Smad2, a major intracellular signaling protein downstream of the TGF-β receptor system (36). Smad2 was examined because a commercial antibody against phospho-Smad2 is available. The parallel arm of TGF-β1 signaling, the Smad3 pathway, was not investigated because an antibody against phospho-Smad3 has not been made available. Treatment with exogenous TGF-β1 markedly increased phospho-Smad2 levels in the podocyte (Fig. 7C). As expected, Smad2 phosphorylation was totally suppressed by the direct TGF-β inhibitor, SB-431542 (Fig. 7C). However, Smad2 phosphorylation was not inhibited by either SU5416 or LY294002 (Fig. 7C).

VEGF comprises a family of potent cytokines that include the VEGF-A class and the less well-described VEGF-B, -C, -D, and -E classes. In the VEGF-A class, differential exon splicing gives rise to at least five different VEGF isoforms (37–39) that have been named according to the number of amino acids they contain. In humans, they are designated VEGF₁₂₁, VEGF₁₄₅, VEGF₁₆₅, VEGF₁₈₉, and VEGF₂₀₆; in mice, the VEGF isoforms have one less amino acid and are called VEGF₁₂₀, VEGF₁₆₄, VEGF₁₈₈, and VEGF₂₀₅ (40). In general, the smaller isoforms (VEGF_{120, 164}) are soluble and freely secreted, whereas the larger isoforms (VEGF_{188, 205}) remain cell associated (41). Nevertheless, the predominant species is VEGF₁₆₅, the most abundantly expressed and secreted isoform (13). The major source of VEGF in the glomerulus is the podocyte (9), and this cell can be stimulated to overproduce VEGF by cytokines such as TGF-β1 (21) and by pathophysiological states such as diabetes (5). The functions of podocyte-derived VEGF are not well established, but recent studies have implicated an important role in the pathogenesis of albuminuria in type 1 and type 2 diabetes (7, 8). It remains unclear, though, how VEGF can influence macromolecular permeability across the glomerular filtration barrier or even what cell type VEGF acts upon.

Several lines of evidence support the existence of a functional VEGF autocrine loop in cultured podocytes. First, exogenous VEGF₁₆₄ exerts a measurable effect on podocyte pathobiology in the form of increased α3(IV) collagen production, a meaningful parameter given the podocyte’s role in GBM assembly. That the enhanced production of α3(IV) collagen was completely blocked by SU5416 attests to the involvement of the VEGF signaling system in mediating α3(IV) collagen expression. Not only did SU5416 completely prevent VEGF-induced α3(IV) collagen, but increasing doses of SU5416 gradually decreased α3(IV) below control levels, indicating a functional role for endogenous VEGF in basal α3(IV) production. The range of SU5416 doses was chosen to encompass the IC₅₀ of ∼1 μmol/l for the inhibition of VEGFR-1 and -2 autophosphorylation but not to infringe on the IC₅₀ of ∼20 μmol/l for the inhibition of platelet-derived growth factor–dependent phosphorylation (15, 32).

That VEGF exerts a substantive effect on podocytes is reinforced by the identification of VEGFR-1, the demonstration of receptor autophosphorylation, and the inhibition of exogenous and endogenous VEGF effects by the receptor inhibitor, SU5416, and by the PI3K inhibitor, LY294002. The presence of VEGFR-1 mRNA and protein was confirmed by RT-PCR, immunoblotting, and fluorescence immunocytochemistry in mouse podocytes. In contrast, VEGFR-2 was undetectable by these means. Our findings extend those of Foster et al. (10) who detected a message for VEGFR-1, but not -2, by RT-PCR in human podocytes. Since treatment with VEGF results in tyrosine phosphorylation of VEGFR-1, signaling through this receptor is potentially responsible for the effects of VEGF on collagen production. It is less likely that the other VEGF receptors known to be present in podocytes (10) can transduce the stimulatory effect of VEGF on α3(IV) collagen production; VEGFR-3 responds only to VEGF-C or -D (42), and neuropilin-1 does not signal by itself (43) but acts more like a coreceptor for VEGFR-2 (14).

Additional proof for the existence of a VEGF autocrine loop in podocytes was obtained by stimulating endogenous VEGF secretion by TGF-β1 (21) and then being able to inhibit an effect of TGF-β1 with anti-VEGF therapy. Both SU5416 and LY294002 reduced the TGF-β1–induced α3(IV) collagen by nearly 70%, suggesting that part of the mechanism for TGF-β1 to increase α3(IV) collagen is mediated by the VEGF system, signaling through the PI3K-Akt pathway. This partial inhibition also implies that TGF-β1 stimulates a component of α3(IV) production that is not mediated by endogenous VEGF or PI3K-Akt signaling. However, the entire TGF-β1 effect on α3(IV) collagen could be inhibited by SB-431542, which blocks all avenues of TGF-β1 signaling by inhibiting the TGF-β type I receptor kinase (21, 33). It should be noted that neither SU5416 nor LY294002 directly inhibited the TGF-β1 signaling pathway, evidenced by lack of effect on Smad2 phosphorylation. On the other hand, treatment with SB-431542 completely abolished Smad2 phosphorylation following TGF-β1 treatment. We conclude that the autocrine VEGF system in cultured podocytes mediates part of the TGF-β1 stimulatory effect on α3(IV) collagen production, possibly through VEGFR-1 and the PI3K pathway.

To further establish the specificity of SU5416 in podocytes, we investigated a cytokine-mediated end point that does not involve the VEGF pathway. TGF-β1 strongly inhibits α5(IV) collagen production (21), whereas VEGF neither increases nor decreases α5(IV) production (Fig. 1B). Therefore, TGF-β1 does not utilize the endogenous VEGF system to suppress α5(IV) collagen. Accordingly, SU5416 fails to modify the inhibitory effect of TGF-β1 on α5(IV) production (Fig. 7B). TGF-β1–mediated suppression of α5(IV) collagen, however, can be reversed by a direct inhibitor of TGF-β signaling, SB-431542 (Fig. 7B). Combined with the prior results on α3(IV) collagen, the fact that SU5416 can inhibit a VEGF-dependent effect of TGF-β1 (e.g., on α3) but not a VEGF-independent effect of TGF-β1 (e.g., on α5) demonstrates that SU5416 specifically targets the VEGF pathway.

The novel role of VEGF in mediating some matrix-promoting effects of TGF-β1 in podocytes is distinct from its classically described properties relating to angiogenesis and vascular permeability and can be added to the expanding repertoire of VEGF actions in normal and pathophysiologic states (44). For diabetic glomerulopathy, a revised pathogenetic scheme can be conceived that incorporates this newly discovered role for an autocrine VEGF loop in promoting podocyte expression of α3(IV) collagen, a principal ingredient of the GBM. The diabetic state gives rise to increased glomerular concentrations of TGF-β1 (45, 46), whether coming from the mesangial cell in response to high glucose (47) or locally increased levels of angiotensin II (48) or from the glomerular endothelium in response to amadori-glycated albumin (49) or leptin (50). Increased levels of TGF-β1 can act in a paracrine manner upon the podocyte, now poised to respond to TGF-β1 because of concordant upregulation of the signaling TGF-β type II receptor (21). In this article, we show that one important consequence of TGF-β system overactivity in diabetic glomerulopathy is to stimulate podocyte secretion of VEGF and the actions thereof, such as increasing α3(IV) collagen production. This effect of TGF-β1 may be amplified by high ambient glucose, which increases both VEGF expression and α3(IV) collagen production in podocytes (21).

Whether stimulation of α3(IV) production by the TGF-β1–VEGF axis contributes to diabetic GBM thickening will need to be evaluated in animal models. In addition, the lack of a role for VEGF in modulating the inhibitory effect of TGF-β1 on α5(IV) collagen may have important consequences. Excess α3(IV) collagen and decreased α5(IV) collagen production may disrupt the orderly assembly of the GBM framework, causing dysfunction of the glomerular filtration barrier. Perhaps VEGF-mediated dysregulation of GBM composition may help to explain why anti-VEGF therapy ameliorates albuminuria in diabetic nephropathy (7, 8).

This work was supported by grants from the National Institutes of Health (DK-61537 to S.C.; DK-44513, DK-45191, and DK-54608 to F.N.Z.; Institutional Training Grant DK-07006 to B.J.; and P30-DK50306 to the Morphology Core of the Gastroenterology Division) and the American Diabetes Association (to F.N.Z.).

The authors thank Dr. Peter Mundel (Albert Einstein College of Medicine) for donating the conditionally immortalized, mouse podocyte cell line, Dr. Michael P. Madaio (University of Pennsylvania) for supplying the human polyclonal anti-α3(IV) collagen antibody, Dr. Jean-Francois Beaulieu and Elizabeth Herring (University of Sherbrooke) for providing the rabbit polyclonal anti-α5(IV) collagen antibody, Martine Kouahou for technical assistance, Dr. Gary P. Swain for help with immunostaining and fluorescent microscopic imaging, and Dr. Amy Wang for her critical reading of the manuscript.