Response of Experimental Retinal Neovascularization to Thiazolidinediones FREE

DIABETIC RETINOPATHY is a major cause of blindness in the United States.¹ Almost all patients in whom diabetes begins before 30 years of age show some evidence of retinopathy 20 years later, which is not surprising, given the finding that the risk for retinopathy is directly related to the degree and duration of hyperglycemia.² In those patients in whom diabetes develops after 30 years of age, retinopathy is less frequent but may be the first manifestation of the disease.² Proliferative diabetic retinopathy (PDR) is the most sight-threatening form of diabetic retinopathy and is characterized by retinal neovascularization. The growth of new vessels on the surface of the retina in PDR can be correlated with the presence of retinal ischemia and increased local expression of vascular endothelial growth factor (VEGF).^{3 - 9} Retinal endothelial cells (RECs) have been shown to express VEGF receptors, and endothelial cells respond to VEGF by activation of multiple signaling molecules, including phospholipase C γ, phosphatidylinositol 3-kinase, protein kinase C, and the extracellular signal-regulated mitogen-activated protein kinase (ERK-MAPK) pathway.^{10 - 13}

A major effort has been made to develop novel therapeutic agents to inhibit retinal neovascularization. Experimental oxygen-induced retinopathy is commonly used in these studies as a model for PDR.^{14 - 15} In this model, VEGF is up-regulated in the inner retina after neonatal mice exposed to hyperoxic conditions are returned to room air, resulting in the development of consistent retinal neovascularization. Several approaches have met with success in this and other ischemic model systems, including those that inhibit VEGF by means of soluble receptor or antibody^{16 - 17} and those that inhibit VEGF receptor activation¹⁸ or its downstream signaling¹⁹ ; however, none of these have been proven efficacious in humans with PDR.

We considered the possibility that a class of drugs currently in use for the primary treatment of diabetes may also inhibit neovascular complications of diabetes independent of their effects on blood glucose level, thus potentially decreasing the risk for PDR for these patients. Thiazolidinediones (TZDs), including troglitazone, rosiglitazone maleate, and pioglitazone hydrochloride, constitute a novel class of drugs that can be used to improve insulin resistance in type 2 diabetes.^{20 - 21} Two members of this class (rosiglitazone and pioglitazone) are currently approved for clinical use. The TZDs are synthetic high-affinity ligands for the nuclear peroxisome proliferator-activated receptor γ (PPAR-γ).^{22 - 24} Expression of PPAR-γ is most abundant in adipose tissue, where it promotes adipocyte differentiation and regulates the expression of genes involved in fatty acid metabolism; however, recent studies have shown a more widespread distribution of PPAR-γ, suggesting that it may affect other tissues.²⁵ Recently, we and others have demonstrated the expression of PPAR-γ on endothelial cells and have shown antiangiogenic effects in corneal and choroidal experimental systems.^{26 - 28}

In the present study, we show that RECs express PPAR-γ protein and that TZDs inhibit angiogenic responses to VEGF in vitro. We then show that 2 TZDs inhibit the development of retinal neovascularization in the oxygen-induced retinopathy model in mice. To further define the mechanism of this effect, we examined the effect of TZDs on expression of VEGF in vivo and VEGF-induced ERK-MAPK phosphorylation in vitro.

We isolated bovine RECs using magnetic beads carrying the endothelium-specific marker Bandeiraea simplicifolia (BS-1; Sigma-Aldrich Corp, St Louis, Mo), as previously described.²⁹ The cells were confirmed to be vascular endothelial cells by means of positive findings of immunostaining for von Willebrand factor (vWF) (Dako, Carpinteria, Calif) and by means of uptake of dil-acetylated low-density lipoprotein (Biomedical Technologies, Stoughton, Mass). Nuclear and cytosolic fractions of protein extracts²⁸ were prepared using the method of Dignam et al.³⁰ Full-length in vitro translated PPAR-γ1 was used as a positive control. Gel electrophoresis and Western blotting were performed as previously described²⁸ using a rabbit polyclonal anti–PPAR-γ antibody (Santa Cruz Biotechnology, Santa Cruz, Calif) (1:1000) and a chemoluminescent detection kit (Lumi-GLO; Kirkegaard & Perry, Gaithersburg, Md). In vitro assays were performed in response to VEGF (10 ng/mL) after 24 hours of serum starvation. Cell proliferation was assayed by means of tritiated thymidine uptake.²⁸ Chemotactic migration was examined using transwell cell-culture chambers (Costar, Cambridge, Mass).²⁸ Capillarylike tube formation was examined by growing RECs in a collagen mixture as previously described.²⁸ The effect of troglitazone or rosiglitazone (at concentrations of 0, 0.1, 1, 10, and 20 µmol/L) added with the VEGF was determined for each of these in vitro assays. Experiments were performed in triplicate and were repeated at least 3 times.

The experiments conformed to the Association of Research in Vision and Ophthalmology Resolution on the Use of Animals in Ophthalmic and Vision Research, and were approved by the University of Southern California Institutional Care and Use Committee, Los Angeles. The reproducible murine model of oxygen-induced retinopathy has been described previously.¹⁵ Briefly, litters of 7-day-old C57BL/6N mice were exposed to a mean (±SD) of 75% ± 2% oxygen for 5 days and then returned to the room air on postnatal day 12. Twelve mice were included in each group (troglitazone- and rosiglitazone-treated and control mice). Eight mice were used for quantitation of retinal angiogenesis or immunohistochemical analysis, and the remaining 4 mice were used for angiography with high-molecular-weight fluorescein dextran. Intravitreal injections of troglitazone or rosiglitazone were performed as described below. Mice of the same age that had been kept in room air were used as controls.

Mice were deeply anesthetized by means of intraperitoneal injection of ketamine hydrochloride (40 mg/kg) and xylazine hydrochloride (10 mg/mL). Intravitreal injections were performed on postnatal days 12 and 14 by delivering 0.5 µL of troglitazone or rosiglitazone maleate (100 µmol/L) diluted in dimethyl sulfoxide (DMSO) to the left eye and DMSO alone to the right eye with a 32-gauge Hamilton needle 200 µm posterior to the limbus.

Mice were deeply anesthetized as described above, and then 100 µL of phosphate-buffered saline solution containing 5 mg of fluorescein isothiocynate–dextran(molecular weight, 2 000 000; Sigma-Aldrich Corp) was injected into the left ventricle. After 1 minute, the mice were killed using an intraperitoneal injection of pentobarbital sodium (75 mg/kg). The eyes were enucleated, and the retinas were dissected and flat-mounted on microscope slides for examination using a fluorescence microscope (BX50; Olympus, Tokyo, Japan).

The mice (postnatal day 17) were killed using an intraperitoneal injection of pentobarbital sodium (75 mg/kg). Enucleated eyes were fixed with 4% paraformaldehyde in phosphate-buffered saline solution and embedded in paraffin. Serial axial sections (3 µm) of the retina were obtained, starting at the optic nerve head. After staining with hematoxylin-eosin, 10 intact sections of equal length, each 30 µm apart, were evaluated. All retinal vascular cell nuclei anterior to the inner limiting membrane were counted in each section; for controls, this value was 0.

Immunohistochemical analysis for the endothelium-specific marker vWF was used to identify intraretinal and preretinal angiogenesis and normal retinal vascular channels. Immunohistochemical analysis for VEGF (sc-507; Santa Cruz Biotechnology) was used to detect its overexpression in the hypoxic retina. Immunoperoxidase detection was performed by the avidin-biotin complex method with 3-amino 9-ethylcarbazole as the red chromogen. Negative controls included omission of primary antibody and use of an irrelevant primary antibody of the same isotype.

The RECs were seeded into 6 multiwell plates in endothelial growth medium(Clonetics, Walkerville, Md). Confluent RECs were starved in Dulbecco's Modified Eagle Medium (DMEM; Cellgro, Herndan, Va) containing 0.1% bovine serum albumin(Sigma-Aldrich Corp) for 8 hours. The RECs, untreated or treated with 25 ng/mL recombinant human VEGF 165 (R & D Systems Inc, Minneapolis, Minn) for 5 minutes, were washed once in cold phosphate-buffered saline solution and lysed in Laemmli buffer (50-mmol/L Tris [pH 6.8], 2% sodium dodecyl sulfate[SDS], and 10% glycerol; BioRad, Hercules, Calif) containing protease inhibitors. Cell lysates were heated to 95°C for 2 minutes, and equal volumes of lysate were subjected to 10% SDS–polyacrylamide gel electrophoresis (PAGE). The blots were incubated with anti–phosphospecific ERK1/ERK2 (p44/p42) antibody (New England Biolabs, Beverly, Mass), followed by incubation with horseradish peroxidase–conjugated secondary antibody (Amersham, Piscataway, NJ). Visualization was performed using the Amersham enhanced chemiluminescence detection system. Lane-loading differences were normalized by reblotting with non–phosphorylation-specific anti-ERK1 antibody (Santa Cruz Biotechnology).

The RECs were then pretreated with vehicle, rosiglitazone maleate (10 µmol/L), or troglitazone (1-20 µmol/L) for 15 minutes before exposure to vehicle alone or VEGF 165 (25 ng/mL) for 5 minutes. Cell lysates were subjected to SDS-PAGE, and the evaluation of p44/p42 ERK-MAPK phosphorylation was performed as already described.

Experimental and control groups were compared using independent sample t tests. Statistical significance was defined as α<.05.

To detect the expression of PPAR-γ protein in bovine RECs, we performed Western immunoblotting of nuclear and cytosolic proteins. Figure 1 shows a band of 52 kd corresponding to PPAR-γ1 that localizes almost exclusively to the nuclear fraction.

Proliferation of RECs was induced using VEGF (10 ng/mL). Rosiglitazone and troglitazone each significantly inhibited VEGF-induced thymidine incorporation, beginning at concentrations of 10 µmol/L (P<.05) and becoming more prominent at concentrations of 20 µmol/L (P<.005) (Figure 2A). Both TZDs were equally effective at this dose range. Viability, as measured by means of trypan blue exclusion, was greater than 95% at all drug concentrations tested.

Rosiglitazone and troglitazone inhibited the chemotactic migration of RECs to VEGF in a modified Boyden chamber in a dose-dependent manner beginning at 10 µmol/L (P<.05) (Figure 2B). Viability was maintained at greater than 95% at all drug concentrations used in this experiment.

Dramatic inhibition of VEGF-induced tube formation by RECs in collagen gels was seen in the presence of rosiglitazone or troglitazone at a concentration of 10 µmol/L (P<.01) (Figure 2C). At a concentration of 20 µmol/L, essentially no tube formation was found. No morphologic evidence of cell death was observed at any of the TZD drug concentrations tested.

To evaluate the antiangiogenic effect of TZDs on oxygen-induced retinopathy, retinas were examined by means of fluorescein-dextran injection angiography. Retinas of the DMSO-injected eyes of animals at postnatal day 17 (5 days of relative retinal hypoxia) contained prominent neovascular tufts extending into the vitreous body at the junction between the perfused and the nonperfused tissue (Figure 3A-B). In contrast, in the retinas from experimental animals treated with 0.5 µL of troglitazone(100 µmol/L) or rosiglitazone maleate (100 µmol/L) and examined on postnatal day 17, the amount of neovascular tissue was markedly reduced, despite the presence of comparable pericentral regions of nonperfusion (Figure 3C). The capillary network and nonperfused areas were well formed, but there was no tuft of neovascularization extending into the vitreous body (Figure 3D). Retinas from normoxic controls demonstrated an evenly distributed stained capillary network (data not shown).

The neovascularization was quantitated histologically by counting the endothelial cell nuclei anterior to the inner limiting membrane. Retinas of DMSO-treated eyes from mice on postnatal day 17 (5 days of relative retinal hypoxia) contained multiple neovascular tufts extending into the vitreous(Figure 4A). These tufts originated from retinal vessels, forming clusters of immature endothelial cells (Figure 4B). Intravitreal injection of troglitazone(Figure 4C-D) or rosiglitazone reduced the histologically evident retinal neovascularization in all 8 animals compared with the DMSO-treated experimental controls (Figure 5). The mean (±SD) number of endothelial nuclei anterior to the inner limiting membrane was significantly reduced from 474 ± 93 in control eyes (DMSO-injected) with oxygen-induced retinopathy to 267 ± 51 in troglitazone-treated eyes (P<.01) and 299 ± 61 in rosiglitazone-treated eyes (P<.01).

Consistent expression of VEGF was seen in all hypoxic retinas, regardless of the agents injected into the vitreous (Figure 6). In DMSO-injected eyes, VEGF was expressed mainly in the ganglion cell layer and to a lesser extent in the inner plexiform layer of the retinas, just beneath the prominent retinal neovascularization (Figure 6A), and adjacent to the intraretinal neovascularization (Figure 6C). Although retinal neovascularization was significantly reduced in troglitazone- or rosiglitazone-injected eyes, VEGF was also prominently expressed in the ganglion cell layer of these TZD-treated retinas (Figure 6B).

After 5 minutes of stimulation, VEGF induced p44/p42 ERK-MAPK phosphorylation(Figure 7). Troglitazone and rosiglitazone(10 µmol/L) inhibited VEGF-induced ERK-MAPK phosphorylation by greater than 50% for p44 ERK1. Dose-response analysis demonstrated that 20-mmol/L troglitazone inhibited ERK-MAPK phosphorylation even more prominently. Troglitazone also eliminated the weak basal p42 ERK-MAPK phosphorylation without affecting cellular shape or attachment. In all experiments, blots of total ERK-MAPK revealed that total ERK-MAPK protein concentration remained unchanged.

Insulin resistance plays a crucial role in the pathogenesis of type 2 diabetes; however, much of the morbidity of this disease results from its complications, including microangiopathy and cardiovascular disease. The TZDs constitute a class of pharmacologic compounds that enhance insulin action through activation of the PPAR-γ receptor.^{22 - 24} These drugs, when used alone or added to traditional oral hypoglycemic agents, result in better control of glycemic markers and amelioration of hyperinsulinemia.^{20 - 21} Recently, there has been considerable interest in the additional benefits of these drugs for other metabolic abnormalities in diabetes, including improvements in lipid profile and blood pressure control.²⁰ Moreover, it has recently been suggested that TZDs independently delay certain diabetic complications through direct activation of PPAR-γ in vascular cells.³¹ Troglitazone inhibits endothelial cell activation^{26 - 27} and platelet aggregation³² and delays development of atherosclerosis in animal models.³³ Beneficial effects on other diabetic complications have also been reported. The TZDs may inhibit microalbuminuria in patients with incipient diabetic nephropathy³⁴ and may protect against nephropathy in Zucker fatty rats.³⁵ Inhibitory effects of TZDs on diabetic neuropathy in streptozocin-induced diabetic rats have also been described.³⁶

In the present report, we provide evidence that TZDs may reduce the complications of diabetic retinopathy. Diabetic retinopathy encompasses the following 5 basic abnormalities: formation of retinal capillary microaneurysms, development of excessive vascular permeability, vascular occlusion, proliferation of new blood vessels, and contraction of accompanying fibrous tissue on the surface of the retina.² Proliferative diabetic retinopathy is particularly devastating and is characterized by the growth of new vessels on the surface of the retina. At the present time, the risk for blindness with untreated PDR is greater than 50% at 5 years, but with appropriate current therapy, it can be reduced to less than 5%.³⁷ Despite these improvements, it remains a very common cause of blindness that would be treated more appropriately by prevention.

Neovascularization is a multistep process that includes degradation of basement membrane and proliferation, migration, and tube formation by endothelial cells.³⁸ The neovascularization process is stimulated by a variety of growth factors and cytokines; however, VEGF has been shown to be central to this process in PDR.^{3 - 8} It has been reported previously that endothelial cells from several sources, including the umbilical vein, aorta, and choroid, express the TZD receptor PPAR-γ; nevertheless, it is important to confirm this finding in the specific endothelial populations being studied, since endothelial cells are heterogeneous in their phenotype and function.^{26 - 28} We show herein that RECs strongly express PPAR-γ1 protein. The PPAR-γ gene produces 2 messenger RNA species by means of alternative promoter use, each containing different 5′ exons that are spliced onto common downstream exons. The PPAR-γ2 protein differs from PPAR-γ1 by the presence of 30 additional amino acids at its N-terminal.³⁹ Although both isoforms of this nuclear receptor are expressed in abundant levels in human adipose tissues, PPAR-γ1 expression is typically much higher than PPAR-γ2 expression in nonadipose tissues.^{25 ,39} Consistent with these findings, RECs express PPAR-γ1 protein that is localized almost exclusively to the nuclear fraction.

The antidiabetic action of the TZDs appears to be mediated primarily through activation of PPAR-γ; however, troglitazone is distinguishable from other TZD PPAR-γ ligands because it also contains an α-tocopherol moiety, which could have effects due to its antioxidant activity⁴⁰ or its ability to inhibit protein kinase C.^{41 - 42} Both TZDs studied in this report inhibited the proliferation, migration, and tube formation of RECs in response to VEGF. There was no apparent difference in the sensitivity of RECs to troglitazone and rosiglitazone, providing strong support for the notion that both drugs have effects on RECs by binding to PPAR-γ.

Previous in vivo studies on the antiangiogenic effects of PPAR-γ ligands have been limited to a VEGF-containing corneal pocket assay²⁷ and a laser injury model of choroidal neovascularization.²⁸ The present study uses a much more pathologically relevant model in which the neovascularization is stimulated by increased VEGF expression in the inner retina. Although the model lacks other metabolic abnormalities found in diabetes, it isolates the VEGF-driven process to the inner retina and allows more precise examination of the mechanism within a clinically relevant microenvironment. By use of qualitative and quantitative analyses, we demonstrated a decrease in the number of microvascular tufts that were induced on the retinal surface, suggesting that the TZDs are inhibiting an early aspect of neovascularization. Again, it was important to demonstrate similar degrees of inhibition of in vivo angiogenesis by 2 PPAR-γ ligands, implying that the drugs are acting similarly through PPAR-γ activation. Although rosiglitazone binds to PPAR-γ with higher avidity than to troglitazone, the TZDs were equally potent in inhibiting VEGF-mediated growth and REC migration. These observations are similar to our previous findings for their antiproliferative and antimigration activity in vascular smooth muscle cells.⁴³ When VEGF expression was examined immunohistochemically, we found that approximately equal amounts of VEGF were expressed in the ganglion cell layer of positive control animals with neovascularization as were found in animals in which retinal neovascularization was inhibited by TZDs, supporting the contention that the TZDs do not interfere with VEGF expression in the region where neovascularization occurs. This contention is further supported by in vitro data demonstrating that hypoxia-induced VEGF secretion by viable retinal pericytes and astrocytes is not significantly reduced by treatment of cells with troglitazone (results not shown).

Previous studies in vascular smooth muscle cells show that PPAR-γ ligands may be acting, at least in part, downstream of growth factor receptor tyrosine kinase activation through inhibition of the ERK-MAPK pathway.^{44 - 45} Although in those studies it was suggested that the effect may be downstream of ERK1/2 activation, we demonstrate here in RECs the significant inhibition of VEGF-induced ERK1/2 phosphorylation by a TZD. Troglitazone has also been shown to inhibit angiotensin II–induced ERK-MAPK activity in vascular smooth muscle cells.⁴⁶ Since the ERK-MAPK pathway plays an important role in cellular proliferation and migration, inhibition of this pathway could explain the effects that we found in RECs in vitro.^{45 - 46}

The TZDs may also have more widespread effects that might improve their effectiveness as inhibitors of neovascularization in PDR. Although VEGF is thought to be the primary effector of PDR, neovascularization occurs in a complex environment of multiple growth factors.³ The ERK-MAPK pathway is known to be a common pathway of activation by multiple growth factors, and PPAR-γ ligands have been shown to inhibit migration mediated by multiple chemoattractants in vascular smooth muscle cells, including insulin-like growth factor I and platelet-derived growth factor; therefore, effects of other growth factors in endothelial cells may be inhibited.⁴⁵ Other possible mechanisms by which PPAR-γ ligands may inhibit retinal vascular dysfunction include inhibition of macrophage activation and inflammation,^{47 - 48} decrease in VEGF receptor expression,²⁷ decrease in matrix metalloproteinase 9 secretion,⁴⁹ decrease in platelet aggregation,³² increase in plasminogen activator inhibitor 1 expression,⁵⁰ induction of endothelial cell apoptosis,⁵¹ and suppression of endothelin 1 secretion from endothelial cells.⁵²

The results of the present study further support the idea that TZDs may have beneficial effects on the diabetic patient beyond those of improving insulin resistance by reducing or delaying the onset of complications such as PDR. The common effect of both agents studied herein suggests that the effect on experimental retinal neovascularization is mediated through PPAR-γ, and our results suggest that much of the effect appears to be mediated downstream of VEGF expression, possibly through inhibition of ERK1 in the ERK-MAPK pathway. Carefully designed clinical studies should be initiated to determine whether diabetic patients currently being treated with PPAR-γ ligands demonstrate inhibition in the development or progression of PDR.

Accepted for publication July 10, 2000.

This study was supported by grants EY01545 (Dr Hinton), EY03040 (Dr Hinton), and HL58328 (Dr Hsueh) from the National Institutes of Health, Bethesda, Md; and by the American Diabetes Association, Alexandria, Va (Dr Law).

The authors would like to thank Laurie LaBree, MS, for assistance in statistical analysis; Christine Spee for culture of retinal endothelial cells; Ernesto Barron for preparation of figures; and Susan Clarke for editorial review.

Corresponding author: David R. Hinton, MD, Department of Pathology, Keck School of Medicine of the University of Southern California, Los Angeles, CA 90033 (e-mail: dhinton@hsc.usc.edu).