Selective Activation of Peroxisome Proliferator–Activated Receptor (PPAR)α and PPARγ Induces Neoangiogenesis Through a Vascular Endothelial Growth Factor–Dependent Mechanism

The WY14643 and the GW1929 compounds were used to specifically activate PPARα and PPARγ, respectively. GW9662 was used to inhibit PPARγ. VEGF was used as positive control. All molecules were from Sigma.

Human umbilical vein endothelial cells (HUVECs) and human dermal microvascular endothelial cells (HDMVECs) (Cambrex) were cultured in EGM2. Migration of endothelial cells was determined using a monolayer denudation assay. Eighty to 90% confluent endothelial cells were scraped with a 1,000-μl pipette tip. Cultures were washed three times with Hanks' balanced salt solution and incubated with EBM2/0.5% fetal bovine serum (FBS). Control cultures received medium alone, whereas experimental groups received PPAR agonists at concentrations of 1 nmol/l, 10 nmol/l, 100 nmol/l, 1 μmol/l, 5 μmol/l, 10, and 20 μmol/l. VEGF was used as positive control at a concentration of 10 nmol/l. The rate of wound closure was determined by photographing cells at three premarked areas at time 0 and at 6 and 12 h. Using a digital imaging system, the rate of cell migration was determined by calculating the difference of the wound area after 6 and 12 h, divided by 2. For cell proliferation experiments, endothelial cells were plated into 96-well plates at 1.0 × 105 or 0.5 × 105 cells/well and allowed to attach overnight. Cells were incubated in EBM2/0.5% FBS with PPAR agonists at concentrations of 1 nmol/l, 10 nmol/l, 100 nmol/l, 1 μmol/l, 5 μmol/l, 10, and 20 μmol/l. VEGF was used as positive control at a concentration of 10 nmol/l. After 24, 36, and 48 h, the amount of proliferation was compared by an MTT assay performed according to manufacturer's instructions (American Type Culture Collection). We also evaluated whether WY14643 and GW1929 were able to increase VEGF levels in endothelial cells using a commercially available ELISA kit (R&D Systems). All of the experiments were performed in triplicate.

We used a commercially available assay (TCS CellWorks), in which proliferating early passage normal human endothelial cells are co-cultured with early passage normal human interstitial cells in 24-well plates, in a specially formulated culture medium. In this assay, cells spontaneously proliferate and then enter a migratory phase, during which they move through the matrix to form, after 11 days in culture, a network of capillary-like tubules (36). This degree of spontaneous angiogenic activity was evaluated and used as a control to quantify the angiogenic effects induced by the addition of PPAR agonists. The following concentrations of WY14643 and GW1929 were used: 1 nmol/l, 10 nmol/l, 100 nmol/l, and 1 μmol/l. VEGF and suramin were used as positive and negative controls, respectively, at the concentration suggested my the manufacturer (TCS CellWorks). On day 11, cells were fixed with ice-cold 70% ethanol, and tubule formation was visualized by immunostaining for the endothelial cell marker CD31 and quantified by image analysis (TCS AngioSys software), as previously described (37,38). Number of tubules, total tubule length, and number of junctions were calculated in untreated cells and in cells stimulated with PPAR-selective agonists, VEGF, and suramin. The ability of WY14643 and GW1929 to increase VEGF levels was evaluated using a commercially available ELISA kit (R&D Systems). Activity of VEGF was inhibited by adding to the culture medium a goat anti-human VEGF-neutralizing antibody (R&D Systems) at a concentration of 500 ng/ml, as previously described (39). All of the experiments were performed in triplicate. Results were expressed as ratio between the angiogenic effect induced by the abovementioned compounds and that observed in untreated cells. All quantifications were performed in a blinded fashion by two independent operators.

Male 8- to 12-week-old C57BL/6J mice were used for these experiments. Mice were anesthetized with an intramuscular injection of ketamine and killed with an overdose of ketamine. A single corneal pocket was created, and pellets containing 0.3 μg WY14643 or GW1929 were prepared and placed in each corneal pocket, as previously described (40). Pellets containing 0.3 μg VEGF were used as positive controls. We also implanted pellets containing 0.03, 0.3, and 3 μg gemfibrozil (Teva Pharma Italia) or rosiglitazone (GlaxoSmithKline) to perform a direct comparative analysis between the angiogenic properties of these two classical PPAR agonists and those of WY14643 and GW1929. Doses were calculated on the basis of the established ability of gemfibrozil and rosiglitazone to bind, respectively, PPARα and PPARγ in vitro, compared with WY14643 and GW1929 (41,42). Pellets containing control buffer were used as negative controls. Eight to 10 eyes were studied in each group. The investigation was in accordance with the A. Gemelli University Hospital Institutional Animal Care and Use Committee. Corneal neoangiogenesis was quantified 6 days after pellet implantation, as established and described in previous reports (40). Vessel length, circumferential extent of neovascularity, number of vessels per cross section, and lumen diameter were evaluated using fluorescence microscopy. For these analyses, mice received an intracardiac injection of 500 μg endothelial cell marker BS-1 lectin, conjugated to fluorescein isothiocyanate (Vector Laboratories). Thirty minutes later, mice were killed, and each eye was enucleated and fixed in 1% paraformaldehyde solution for 1 h. Corneal hemispheres were prepared under the dissecting microscope, placed on glass slides, and analyzed using a fluorescent microscope. To verify that the neoangiogenic effects of WY14643 and GW1929 were specifically mediated by PPARα and PPARγ, respectively, the experiments were repeated in mice lacking the PPARα gene (PPARα^−/− mice, homozygous for the Pparatm1Gonz gene) (Jackson Laboratories) and in mice in which PPARγ activity was blocked by the specific inhibitor GW9662 (Sigma) (43,44). In addition, the ability of PPARα^−/− mice and mice treated with the PPARγ inhibitor GW9662 to respond normally to a prototypical angiogenic stimulus were to tested by implanting pellets containing VEGF. All quantifications were performed in a blinded fashion, by two independent operators.

We analyzed whether PPARα- and PPARγ-induced corneal neoangiogenesis occurred in association with production of VEGF. Levels of VEGF were studied at the RNA level by RT-PCR and at the protein level by ELISA and immunohistochemistry. For this set of experiments, the right corneas were implanted with pellets containing WY14643 or GW1929, whereas the left corneas of the same animals were implanted with pellets containing control buffer and served as internal control. For RT-PCR analyses, corneas were harvested 4 days after pellet implantation, and RNA was extracted as previously established (45). cDNA was obtained and amplified using the SuperScript preamplification system (Gibco-BRL). Signals were normalized to 18s rRNA using optimal 18S primer–to–competimer ratios as determined for the target gene following the manufacturer's recommendations (Ambion). The following primer pairs and PCR conditions were used: 5′-CACATCTGCAAGTACGTTCGTTT-3′ and 5′-GTTCAGAGCGGAGAAAGCATTTG-3′ with 30 cycles of 94 (30 s), 62 (1 min), and 72°C (1 min). Results were expressed as mRNA fold increase, calculated as the ratio between VEGF mRNA expression in right and left (control) corneas. For ELISA, corneas were harvested 6 days after pellet implantation. VEGF₁₆₅ levels were measured using a commercially available kit (R&D Systems), as previously described (45). Results were expressed as protein fold increase, calculated as the ratio between VEGF protein levels in right and left (control) corneas. For immunohistochemical staining, eyes treated with pellets containing PPARα and PPARγ synthetic agonists were excised 6 days after pellet implantation and fixed in 1% paraformaldehyde solution for 1 h. After fixation, corneas were embedded in paraffin and cut in cross sections. VEGF immunostaining was performed using a rabbit polyclonal anti-mouse VEGF antibody (Santa Cruz Biotechnology) followed by a biotinylated goat anti-rabbit immunoglobulin secondary antibody (Signet Labs). Negative control slides were prepared by substituting preimmune rat serum.

For immunoblotting, homogenates of corneal tissues were analyzed. Proteins (40 μg/lane) were separated in 10% SDS-polyacrylamide gels and transferred onto polyvinylidine difluoride membranes. Membranes were incubated with antibodies against eNOS (1:1,000) (BD Biosciences Pharmingen), phospho-eNOS (Ser1177, 1:1,000) (Cell Signaling Technology Company), phospho-Akt (Ser473, 1:500) (Cell Signaling Technology Company), and Akt (1:1,000) (New England Biolabs). Antibody binding was detected with horseradish peroxidase–conjugated secondary antibodies (1:2,000) (Chemicon) and enhanced chemiluminescence system (GE Healthcare Bioscience). Then, the blots were reprobed with total eNOS (1:1,000) (Transduction Laboratories), Akt (1:5,000) (Sigma), or actin (1:5,000) (Sigma).

Activity of VEGF was systemically inhibited in vivo by transfection of mice thigh muscles with a plasmid DNA encoding a soluble form of the murine VEGF receptor Flt-1, as previously described (45). The plasmid was provided by Dr. Kensuke Egashira. Soluble Flt-1 can suppress VEGF activity both by sequestering VEGF and by functioning as a dominant-negative inhibitor of VEGF receptors. Control mice received an equal amount of empty plasmid with an intramuscular injection on the same time schedule. Eight eyes were analyzed in each group.

All results are expressed as means ± SE. Differences were analyzed by Student's t test and considered statistically significant at P < 0.05.

We first tested whether PPARα and PPARγ activation stimulates endothelial cell migration or proliferation, two events that are crucial for angiogenesis. No increased migration was seen in HUVECs treated with PPAR agonists (Fig. 1A). Likewise, PPAR agonists were unable to induce proliferation of HUVECs, compared with control (Fig. 1B). Similar results were obtained using a microvascular cell line (HDMVECs) (Fig. 1C and D). PPARα and PPARγ agonists were also unable to increase VEGF production, measured by ELISA, in endothelial cells (data not shown).

For these experiments, we used a co-culture system containing both endothelial cells and interstitial cells (TCS CellWorks). In this assay, endothelial cells spontaneously migrate and proliferate to form tubule-like CD31-positive structures (Fig. 2A). Treatment with the anti-angiogenic molecule suramin resulted in abolishment of tubule formation (Fig. 2B). In contrast, treatment with the angiogenic cytokine VEGF led to significant increase of tubule formation (Fig. 2C). Interestingly, both PPARα and PPARγ agonists exhibited the ability to strongly increase tubule formation (Fig. 2D and E). The neoangiogenic properties of these compounds were quantified in terms of number of tubules, tubule length, and number of junctions and compared with those displayed by untreated cells and VEGF-treated cells (Fig. 2F). PPARα and PPARγ agonists induced significant increase of tubule number, tubule length, and number of junctions, with the maximum effect being observed at the dose of 100 and 10 nmol/l, respectively. Protein analyses also demonstrated that, in culture media of cells treated with PPARα and PPARγ agonists, VEGF levels were significantly higher than in untreated cells (data not shown).

The murine corneal model of angiogenesis was used for these experiments. Pellets containing selective PPAR agonists were implanted in the corneas of mice, and the resulting neoangiogenic response was analyzed 6 days later. BS1-lectin fluorescent staining demonstrated an important neoangiogenic response in eyes treated with pellets containing 0.3 μg WY14643 or GW1929 (Fig. 3A and B). In contrast, no angiogenic response was observed using pellets containing gemfibrozil or rosiglitazone (Fig. 3C and D). As expected, VEGF pellets induced strong neoangiogenic growth (Fig. 3E), whereas no angiogenic response was obtained in corneas implanted with pellets containing control buffer (Fig. 3F). To confirm that the neoangiogenic effect of WY14643 was mediated by PPARα, WY14643-containing pellets were implanted in the corneas of PPARα^−/− mice. No angiogenic response was detected in these animals (Fig. 3G). Similarly, the PPARγ-specific activity of GW1929 was confirmed by implanting GW1929-containing pellets in the corneas of mice that had been pretreated with the specific PPARγ inhibitor GW9662. No angiogenic response was detected in these animals (Fig. 3H). Importantly, both PPARα^−/− mice and mice treated with the specific PPARγ antibody responded normally to stimulation with pellets containing VEGF (Fig. 3I and J), demonstrating that these animals do not have an endogenous impairment of angiogenesis.

To quantify the extent of PPARα- and PPARγ-induced corneal angiogenesis, measurements of vessel length, circumferential neovascularity, number of vessels per cross section, and lumen diameter were carried out. These analyses demonstrated that stimulation of the PPARα and PPARγ pathways resulted in a neoangiogenic process that was statistically similar, in terms of morphological characteristics, to that induced by VEGF (Fig. 4). The average length of neovessels was 0.73 ± 0.11 mm in PPARα-induced neovascularization, 0.71 ± 0.12 mm in PPARγ-induced angiogenesis, and 0.75 ± 0.09 mm in the VEGF group. Mean lumen diameter was 8.02 ± 1.20 μm in PPARα-induced neovascularization, 7.73 ± 1.10 μm in the PPARγ group, and 8.33 ± 1.10 μm in the VEGF group. Circumferential extent of corneal neovascularity was 102.30 ± 7.50° in PPARα-induced neovascularization, 99.10 ± 3.90° in the PPARγ group, and 107.20 ± 4.40° in the VEGF group. The number of vessels per cross section was 121.30 ± 6.90 in PPARα-induced neovascularization, 112.30 ± 7.20 in PPARγ-induced angiogenesis, and 131.21 ± 7.40 in the VEGF group.

First, we evaluated whether VEGF is expressed in association with PPARα- and PPARγ-induced neovascularization. RT-PCR and ELISA analyses demonstrated that VEGF RNA and protein levels are significantly increased in corneas treated with WY14643 and GW1929 (Fig. 5A and B). Second, we performed immunohistochemical analysis that showed that corneas treated with PPAR agonists were strikingly immunopositive for VEGF (Fig. 5C and D). VEGF-producing cells during corneal neovascularization were immunopositive for vimentin (data not shown), a fibroblast mesenchymal marker, suggesting that PPAR agonists stimulate expression of endogenous VEGF in interstitial fibroblasts within the neovascular foci. No VEGF-positive staining was seen in corneas implanted with control pellets or where control staining was performed (Fig. 5E and F). We also investigated the potential impact of PPAR activation on other important angiogenic players, such as eNOS and Akt. We found that, in corneas implanted with pellets containing WY14643 and GW1929, there is a significant upregulation of eNOS phosphorylation at Ser-1177 and Akt phosphorylation at Ser-473 (Fig. 6). Taken together, these findings indicate that a mechanism involving VEGF, eNOS, and Akt contributes to PPARα- and PPARγ-induced angiogenesis in vivo.

Following the observation that PPARα- and PPARγ-induced endothelial tube formation in vitro and neoangiogenesis in vivo occurs in association with VEGF production, we tested the hypothesis that the angiogenic properties of PPAR agonists might depend on VEGF activity. By adding an anti-human VEGF-neutralizing antibody to the medium of the endothelial cell/interstitial cell co-culture assay, we found that inhibition of VEGF activity was sufficient to significantly decrease PPARα- and PPARγ-induced tubulogenesis (Fig. 7A and B). In addition, we suppressed VEGF activity in vivo and evaluated whether pellets containing WY14643 or GW1929 were still able to induce corneal angiogenesis. In vivo inhibition of VEGF was accomplished using the soluble Flt1 plasmid, which suppresses VEGF activity both by sequestering VEGF and by functioning as a dominant-negative inhibitor of VEGF receptors (45). Mice transfected with the empty plasmid were used as controls. A dramatic reduction of PPARα- and PPARγ-induced corneal neoangiogenesis was observed when VEGF activity was suppressed (Fig. 7C–F). Quantification analyses demonstrated that inhibition of VEGF activity resulted in statistically significant reduction of PPARα- and PPARγ-induced neoangiogenesis, measured in terms of vessel length, circumferential extent of neovascularity, and number of vessels per cross section (Fig. 7G). These findings demonstrate that selective PPARα and PPARγ agonists induce angiogenesis via a VEGF-dependent mechanism.

In recent years, the angiogenic potentials of TZDs have been intensively investigated. Some studies have reported that ciglitazone, troglitazone, and rosiglitazone suppress migration, proliferation, and differentiation of endothelial cells and inhibit angiogenesis in a number of experimental models (17–21). Other studies have instead provided evidence that the same molecules may have important pro-angiogenic effects. In particular, it has been shown that TZDs are able to increase VEGF expression in smooth muscle vascular cells and macrophages (14,15), that rosiglitazone increases number and migratory activity of endothelial progenitor cells (16) and promotes angiogenesis after focal cerebral ischemia in rats (22), and that troglitazone induces the expression of VEGF and its receptors in cultured cardiac myofibroblasts (35). In all of these studies, the authors have assumed that the effects induced by TZDs were mediated by the activation of PPARγ without considering that TZDs are nonselective and nonspecific ligands of this nuclear receptor, because they are able to stimulate several PPARγ-independent pathways that are potentially important in angiogenesis (27–33,46). Similarly, the role of PPARα in angiogenesis has been investigated using fibrates, which not only activate PPARα, but also bind other nuclear and superficial receptors and regulate important PPARα-independent pathways (23–26). Therefore, to conclude that the effects of TZDs and fibrates on angiogenesis specifically depend on PPAR, all of the potential activities that these molecules exert on pathways that are PPAR independent should be excluded.

In this study, we used selective PPARα and PPARγ synthetic agonists and found that they were able to enhance endothelial tube formation in vitro and induce neovascularization in vivo. We also demonstrated that these effects were specifically PPAR mediated, by showing that the PPARα activator WY14643 was unable to induce angiogenesis in mice lacking the PPARα gene and that the PPARγ activator GW1929 did not stimulate angiogenesis in mice treated with a PPARγ-specific inhibitor. Importantly, PPARα^−/− mice and mice treated with the PPARγ inhibitor showed a normal response on stimulation with VEGF, demonstrating that these animals do not have an endogenous impairment of angiogenesis and providing additional confirmation that WY14643 and GW1929 inability to induce angiogenesis in these mice depends on PPAR lack of function. Another interesting finding was that gemfibrozil and rosiglitazone, two prototypical fibrates and TZDs that are generally viewed as classical PPAR agonists and have been widely used to respectively test the biological effects of PPARα and PPARγ, were unable to induce angiogenesis in our experimental setting. This finding might help us to understand the controversial results in the literature about the angiogenic properties of PPARs. In this respect, it is important to point out that this is the first time that the angiogenic effects of TZDs and fibrates are directly compared with those of synthetic PPAR-specific agonists, such as WY14643 and GW1929. These data clearly eliminate the possibility that our observations differ from those of other research groups as a result of different experimental settings and/or surgical procedures. They instead support the concept that fibrates and TZDs cannot be considered specific and selective PPAR agonists and that not all of their biological activities can be attributed to activation of PPARs.

Another important finding of this study is the demonstration that PPARα- and PPARγ-induced angiogenesis does not occur through direct stimulation of endothelial cell migration or proliferation. In contrast, PPARα and PPARγ agonists stimulate angiogenesis indirectly, through upregulation of the angiogenic cytokine VEGF. The in vivo neovascularization resulting from the activation of PPARα and PPARγ is morphologically similar to that induced by VEGF and is associated with local VEGF production. Most importantly, suppression of VEGF activity is sufficient to inhibit PPARα- and PPARγ-induced angiogenesis both in vitro and in vivo. However, it is important to note that VEGF inhibition does not completely suppress PPARα- and PPARγ-induced neovascularization. Therefore, it is possible to speculate that additional cytokines might contribute to the angiogenic process stimulated by PPARs. In this scenario, we also show that PPARα- and PPARγ-induced neovascularization occurs in association with enhanced phosphorylation of eNOS and Akt. This finding is important because NO and VEGF display reciprocal regulatory activities that are crucial for angiogenesis, with the fundamental contribution of Akt. On one hand, VEGF may activate eNOS by activating Akt. On the other hand, NO can mediate VEGF expression by stimulating hypoxia-inducible factors and heme oxygenase-1 activity. Although the precise molecular mechanisms responsible for PPAR-induced angiogenesis still merit further investigation, our data provide new insights to understanding the ability of PPARs to modulate the angiogenic process in vivo.

The discovery that both PPARα and PPARγ are able to induce angiogenesis also suggests the possible existence of a synergistic interaction between the two receptors, although we have no data regarding their possible relative roles in the activation of VEGF and/or other angiogenic molecules. This concept might be particularly important when considering the biological activities of PPARα/γ dual agonists. In addition to reducing glucose levels and improving the lipid profile, these molecules also display beneficial activities on cardiac and endothelial function, oxidative stress, and atherosclerosis (47). It is possible to hypothesize that these effects might partially result from the ability of PPARα/γ dual agonists to induce angiogenesis and VEGF production. Similar mechanisms might be hypothesized to explain the potential carcinogenic effects of some PPARα/γ dual agonists (48) and their ability to induce edema and precipitate heart failure (49,50).

Although our data need to be confirmed in other experimental models and in humans, the observation that neoangiogenesis is a potential biological result of PPAR activation might help to understand, at least in part, some of the biological and clinical effects of TZDs and fibrates. For instance, PPARγ-induced VEGF upregulation might be responsible for the recently reported ability of rosiglitazone to ameliorate endothelial dysfunction in type 2 diabetes (51) and increase number and migratory activity of endothelial progenitor cells (16). Because VEGF augments vascular permeability (52), increased production of this cytokine might also contribute to edema, a common side effects of TZDs. It is also possible that PPARγ-induced neoangiogenesis has deleterious effects in subjects with type 2 diabetes. Although angiogenic therapy has been widely regarded as an attractive approach both for treating ischemic heart disease and for enhancing arterioprotective functions of the endothelium, a variety of studies have also suggested that neovascularization contributes to the growth of atherosclerotic lesions and is a key factor in plaque destabilization and rupture (53). These negative effects might contribute to the recently reported potential adverse cardiovascular effects of rosiglitazone in type 2 diabetes (54). Similarly, some clinical activities of fibrates, such as preservation of renal function and amelioration of endothelial dysfunction (55,56), might be partially explained by PPARα-induced angiogenesis and VEGF production.

In conclusion, we used selective synthetic agonists of PPARα and PPARγ and demonstrated that the stimulation of these nuclear receptors results in the activation of a strong neoangiogenic process in vitro and in vivo. This angiogenic response does not occur through direct stimulation of endothelial cells but is instead dependent on VEGF activity. These findings provide new information to understand the biological, clinical, and therapeutic effects of drugs that stimulate the activity of PPARα and PPARγ, with potentially important implications for the management of subjects affected by type 2 diabetes.