Stimulation of Neovascularization by the Anti-angiogenic Factor PEDF

Abstract

purpose. Examine the effect of (pigment epithelium–derived growth factor; PEDF) on laser-induced choroidal neovascularization (CNV).

methods. Adult C57Bl/6 mice were anesthetized and four laser spots were placed in each quadrant of the fundus with a krypton red laser (614 nm, 50 μm, 0.05 second, 200 mW). Animals were treated with various doses of PEDF administered with miniosmotic pumps implanted subcutaneously. Seven days after laser treatment, mice were perfused with 3% FITC high-molecular-weight dextran, the eyes enucleated, and neovascularization analyzed by confocal microscopy. Data were recorded as the volume of the neovascular complex. The effect of PEDF on endothelial cell migration, vascular tube formation in synthetic basement membrane, and VEGF production was also determined.

results. Mice receiving a lower dose of PEDF (90 μg/mL) had significantly decreased areas of CNV. A high dose of PEDF (360 μg/mL) significantly increased CNV, whereas an intermediate dose (180 μg/mL) of PEDF had no effect. PEDF inhibited endothelial cell migration and vascular tube formation at lower doses (0.5–5 μg/mL). High doses of PEDF (25–50 μg/mL) stimulated endothelial cell migration, enhanced vascular tube formation in vitro, and stimulated VEGF production from endothelial cells. Neutralizing anti-VEGF antibody completely reversed the stimulatory effects of high doses of PEDF on CNV in vivo.

conclusions. PEDF demonstrates opposing effects on CNV and endothelial cell function. Whereas low doses are inhibitory, high doses can augment the development of the neovasculature. These results suggest that the effects of PEDF on neovascularization are more complex than originally believed and that caution should be exercised when PEDF therapies are considered.

Neovascularization is a major cause of vision loss in patients with age-related macular degeneration (AMD), diabetic retinopathy, and retinopathy of prematurity (ROP). Neovascular AMD is the aggressive variant of AMD characterized by abnormal, new vessel growth into the subretinal space from the underlying choroid (i.e., subretinal or choroidal neovascularization [CNV]), leading to severe visual loss in these patients. This complication is associated with the atrophy and senescence of retinal pigment epithelial (RPE) cells and microfractures in Bruch’s membrane, as well as exudation of fluid and hemorrhage into the subretinal space resulting in detachment of the overlying neurosensory retina and damage to the retinal photoreceptors. Consequently, it is important to gain insights into the mechanisms that regulate CNV. 

Recent evidence has suggested that natural inhibitors of angiogenesis regulate developmental angiogenesis and prevent abnormal neovascularization in adults. Thus, these agents have been considered as therapy for ocular disorders and cancer. ^{1 2 3 4} Many of these compounds help maintain quiescence of the normal vasculature. Stimulation of new vessel growth in adults appears to be a consequence of an imbalance between these agents and stimulators such as vascular endothelial growth factor (VEGF). ^{5 6} A major natural inhibitor in the vitreous and cornea of the eye is pigment-epithelium–derived growth factor (PEDF). ^{4 6 7} This protein is produced by RPE and is found in high concentration in the retina, vitreous, and cornea. PEDF has been shown to induce apoptosis in endothelial cells and prevent migration of these cells in vitro. ⁶ In addition to its antiangiogenic properties, PEDF appears to have neuroprotective properties. ⁸ It can induce neurite outgrowth in cultured retinoblastoma, ⁹ it supports survival of photoreceptors, ¹⁰ and it protects neurons from apoptosis in vitro and in vivo. ^{11 12} Thus, PEDF can display opposing effects that may depend on the target cell type. 

Recent studies have examined the ability of PEDF to inhibit angiogenesis in several models of ocular disease. ^{4 13 14 15} Promising studies in animal models have suggested that PEDF may be an effective treatment for patients with AMD. Recently, a phase I dose-escalation trial studying the effect of increasing concentrations of adenoviral PEDF injected into the vitreous cavity in a subset of patients with advanced exudative AMD was initiated. ¹⁶ This is the first documented gene therapy trial for an ocular disease in humans. We have examined the effect of PEDF on neovascularization in a laser model of CNV. Our results demonstrate that PEDF can inhibit or promote angiogenesis, depending on the dosage. Furthermore, in vitro studies demonstrate that PEDF can also have both inhibitory and stimulatory effects on cultured endothelial cells. Our results suggest that the effects of PEDF on angiogenesis are highly complex and caution should be exercised as human therapy with this molecule is considered. 

Materials and Methods

Mice

C57Bl/6 (B6) mice were purchased from the National Cancer Institute (Frederick, MD). The authors confirm adherence to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Cell Lines

Human dermal microvascular endothelial cells (HMVECs) were obtained from Cambrex, Inc. (Walkersville, MD) and maintained in endothelial cell basal medium (EBM; Cambrex, Inc.) with 10% FCS and used at passages 4 to 8. 

Laser-Induced CNV

A krypton laser was used to induce choroidal neovascularization (CNV) in the eyes of C57/BL6 wild-type mice. ¹⁷ Four laser spots were placed in each fundus (n = 5 in each group) in the peripapillary area with a krypton red laser (614 nm, 50 μm, 0.05 second, 200 mW). After 7 days, the animals were perfused with 3% FITC-conjugated high-molecular-weight dextran (2000 kDa). After the eyes were enucleated, sclerochoroidal wholemounts were created. Bruch’s membrane was stained with a Cy-3–conjugated anti-elastin antibody. CNV was identified as FITC-perfused vessels above the plane of Bruch’s membrane on confocal microscopy, and images were captured in a three-dimensional stacked manner to facilitate volumetric analysis. Images were analyzed by computer (Metamorph; Universal Imaging, West Chester, PA). Data are recorded as volume of CNV (in cubic micrometers) ± SE. All CNV volumes for all eyes in a single treatment group were averaged and compared individually with control preparations, by using Student’s t-test. 

Reagents

Rabbit anti-PEDF peptide antibody was generated as described. ¹⁸ PEDF protein was purified from a human embryonic kidney (HEK) 293 carcinoma cell line stably transfected with a plasmid containing the human PEDF open reading frame under the direction of the CMV promoter. PEDF protein was fractionated by cation-exchange fast protein liquid chromatography (FPLC; Amersham Pharmacia Biotech, Piscataway, NJ), as described elsewhere. ¹⁸ Two PEDF species were recovered from FPLC, each of which was found to be homogeneous (greater than 99% pure) on electrophoresis. The major PEDF species, which comigrated with PEDF from human vitreous samples, was used for the experiments in this study. PEDF was administered by the use of a miniosmotic pump (Alzet; Durect Corp., Cupertino, CA), which holds a total volume of 200 μL. The pumps were implanted subcutaneously to allow continuous delivery of the protein over the course of the experiment. Delivery was based on diffusion; however, the amount delivered per hour was known (1 μL/h). Concentrations are reported as micrograms per milliliter in all figures. Neutralizing anti-mouse VEGF antibody was purchased from R&D Systems (Minneapolis, MN). Neutralizing anti-PEDF antibody was provided by a coauthor (ED). 

Synthetic Matrix Assay

Growth of human vascular endothelial cells (Clonetics, San Diego, CA) in a three-dimensional synthetic matrix (Matrigel; BD Biosciences, New Bedford, MA) was performed as has been described. ¹⁷ The matrix was prepared according to the manufacturer’s instructions, by diluting 1:2 with serum free EBM. Gels were prepared in 48-well culture plates with 250 μL gel covered with 0.5 mL EBM. Cells (2 × 10⁴) were seeded on the surface of the gel and incubated for 8 hours at 37°C in 5% CO₂. 

Endothelial Cell Migration

In vitro migration assays were performed in a modified Boyden chamber as described. ¹⁹ Bovine adrenal capillary endothelial cells were grown in Dulbecco’s modified Eagle’s medium, and 10% donor calf serum (and 1% endothelial cell mitogen; Biomedical Technologies, Inc., Stoughton, MA). Cells were serum starved overnight, harvested, resuspended in control medium (DMEM+0.1% BSA), plated in an inverted chamber at 3 × 10⁴ cells/well on the bottom side of a gelatinized 5-μm porous membrane (Nucleopore Corp., Pleasanton, CA), and incubated at 37°C for 2 hours to allow attachment. The chambers were reinverted, samples placed in the top wells, and migration allowed for 4 hours. Membranes were fixed, stained (DiffQuick, Fisher Scientific, Fair Lawn, NJ), and mounted, and the number of cells migrating to the top side of the membrane was counted in 10 high-power fields. Basal migration (0.1% BSA) is used as a negative control, migration toward bFGF (10 ng/mL), as a positive control. Samples were tested in quadruplicate, and experiments were repeated at least three times. 

VEGF Production in Endothelial Cells

Human endothelial cells were grown in EBM with 10% FCS. They were plated at 5 × 10⁵ cells/mL in serum-free medium overnight. Various concentrations of PEDF were then added, and the supernatants were harvested 24 hours later. VEGF concentration was determined by ELISA, using a VEGF anti-human ELISA kit (R&D Systems). Results are expressed in picograms of VEGF per milligram total protein. 

Results

Our initial studies using PEDF protein to treat CNV involved injection directly into the vitreous cavity on the day after laser treatment. Contrary to our expectations, we found that concentrations as low as 500 ng/mL stimulated new vessel formation (data not shown). Consequently, we decided to use a subcutaneously implanted pump to deliver PEDF to circumvent any effects on angiogenesis as a result of iatrogenic trauma. This method has the advantage of delivering uniform concentrations of the protein over the entire experimental period without additional insult to the eye. As shown in Figure 1 , the effect of PEDF was dose dependent, with two opposing effects noted on the extreme ends of the curve. When PEDF was administered at a lower dose (90 μg/mL), it significantly reduced the volume of the neovascular complex (Fig. 1) . However, at a higher dose, we observed the loss of inhibitory effect (180 μg/mL). A further elevation in the dose of PEDF significantly increased the appearance of new vessels (360 μg/mL). Figure 2A shows sample neovascular complexes obtained by confocal microscopy in an untreated mouse (Fig. 2A) , as well as inhibition with 90 μg/mL PEDF (Fig. 2B) , no effect at 180 μg/mL PEDF (Fig. 2C) , and augmentation at 360 μg/mL PEDF. 

One possible explanation for the effect of high does of PEDF on CNV may be that PEDF activates inflammatory cells. Inflammation has been shown to be involved in the evolution of the CNV in this model. ^{20 21} We have been unable to demonstrate an effect of PEDF on the activation of macrophages, dendritic cells, or T cells (data not shown). 

The use of recombinant PEDF prepared from cell lines raises the possibility that the augmented neovascularization was due to a contaminant in the preparation. This contaminant would have no effect at low doses but would be revealed when the PEDF dosage is increased. We tested this possibility by treating mice with PEDF alone or in the presence of an antibody neutralizing to the protein. Figure 3 shows that all the augmenting capacity was neutralized by the anti-PEDF antibody (360 μg/mL), either in the same or separate pumps. This suggests that the augmentation observed is the result of the activity of PEDF and not a contaminant. 

We further characterized the effects of PEDF in vitro with two well-established assays for endothelial cell function, endothelial cell migration, and vascular tube formation. As shown in Figure 4 , lower doses of PEDF (0.5 and 5.0 μg/mL) inhibited the endothelial cell migration induced by bFGF. When the concentration of PEDF was raised to 25 μg/mL, migration in response to bFGF was only partially inhibited, whereas the 50-μg/mL concentration of PEDF did not inhibit migration. When PEDF was used in the absence of bFGF, it induced migration of endothelial cells at the higher concentrations of 25 and 50 μg/mL. Thus, PEDF can stimulate endothelial cell migration in the absence of growth factors. A similar effect of PEDF was observed when vascular tube formation in synthetic matrix was examined. Figure 5 shows that 5 μg/mL PEDF inhibited tube formation, whereas 25 and 50 μg/mL did not. Thus, in two assays for endothelial cell function, the dual dose-dependent effect of PEDF was evident. 

VEGF is the primary mediator of angiogenesis is in the eye. ^{22 23 24} Consequently, we explored the ability of PEDF to stimulate VEGF production in endothelial cells. Figure 6 shows that high concentrations of PEDF stimulated VEGF production by HMVECs. Thus, one explanation for our results is that high concentrations of PEDF stimulate VEGF production by endothelial cells, which promotes angiogenesis. We tested this in vivo in the experiment shown in Figure 7 . Administration of anti-VEGF antibody (360 μg/mL) in a separate, subcutaneous pump at the same time as the delivery of high-dose PEDF (360 μg/mL) resulted in a complete reversal of high dose PEDF-mediated angiogenesis. Doubling the concentration of anti-VEGF (720 μg/mL) inhibited both the PEDF-induced angiogenesis as well as the CNV induced by the laser in the presence or absence of PEDF. These results are consistent with the idea that high doses of PEDF stimulate excess VEGF production, which leads to increased CNV. 

Discussion

The importance of angiogenesis in pathologic processes such as tumor formation and eye disease has led to the search for viable therapies that target new blood vessel formation. Recent studies have focused on naturally occurring inhibitors found in normal tissues, where they help maintain a quiescent vasculature. Compounds such as thrombospondin, ³ endostatin, ² angiostatin, ¹ and PEDF ^{4 7} are broad-spectrum inhibitors that can target vascular endothelium activated by numerous angiogenic stimuli and inhibit angiogenesis by inducing apoptosis. 

Recent studies on ocular angiogenesis have focused on PEDF as a potential therapy for retinal and choroidal angiogenesis. ^{4 13 14} PEDF is a natural component of ocular tissues that has shown promise in inhibiting blinding complications of neovascular diseases of the eye. ^{7 25} In a mouse model of ROP, PEDF significantly inhibited retinal angiogenesis. ^{4 18} In a laser model of CNV, intraocular delivery of PEDF in an adenoviral vector reduced angiogenesis beneath the retina. ^{13 14} These promising studies have led to the development of a phase I clinical trial involving delivery of PEDF by an adenoviral vector. ¹⁶ Patients with AMD are given an escalating dose of Ad(GV)PEDF by intravitreal injection. 

We examined the effect of soluble PEDF on CNV by using a continuous-delivery miniosmotic pump. Our results show that there are opposing effects of this protein on laser-induced CNV. A relatively low dose of PEDF (90 μg/mL) over the treatment period was shown to inhibit angiogenesis, as would be predicted from previous studies with this molecule. ^{13 14} However, the intermediate dose of 180 μg/mL had no effect, and as the dose was further raised (360 μg/mL), the new vessel formation significantly increased. These results were reflected in two in vitro assays for endothelial cell functions thought to be critical for angiogenesis, migration, and morphogenesis (tube formation). In these assays, higher doses of PEDF augmented endothelial cell migration in the presence of bFGF. Moreover, when tested alone, PEDF was capable of inducing migration of endothelial cells at high concentrations. A possible explanation for these results was obtained when we observed that high doses of PEDF can stimulate VEGF production by endothelial cells and was supported by the finding that anti-VEGF antibodies completely reversed the PEDF-mediated stimulatory effect on CNV. 

The opposing effects of PEDF suggest that there may be two receptors for this protein that differ in their affinity for the ligand. The one with higher affinity, once engaged, triggers a signaling cascade that results in apoptosis, as was indeed observed in the activated endothelial cells. ⁶ The low-affinity receptor may be stimulatory or produce a survival signal. Unfortunately, the receptor(s) for PEDF is (are) not known, but studies have suggested that there may be receptors in the retina. ^{26 27} The well-defined neurotrophic effects, which appear to generate survival signals and the anti-angiogenic effects, support the two-receptor idea. In addition, it has been suggested that the antiangiogenic activity and neurotrophic activities reside in separate regions of the molecule, also suggesting more than one receptor. ²⁸ This must be studied further. 

A second mechanism may be related to the ability of PEDF to bind the extracellular matrix. ²⁹ Perhaps these interactions can induce conformational changes in PEDF and result in different physiological outcomes when a receptor is engaged. Interaction with different ECM components at different concentrations may account for different activities. PEDF also has the ability to bind glycosaminoglycan. This region has been reported to be separate from the neurotrophic site. ²⁸ Binding to different GAG proteins on the cell surface over a range of concentrations would help account for the observed effects. It should be noted that pro- and antiangiogenic effects of the serpin PAI-1 have been observed. ³⁰  

A recent study examined the effect of PEDF protein in the mouse ROP model. ⁴ These investigators showed a stunning effect on the development of retinal vessels in this model. They did not observe a biphasic effect, as we saw in the CNV model. Although we do not have a complete explanation for these differences, we believe there are several possibilities. First, we have always attained such levels of inhibition in the ROP model with the preparations of PEDF we used. ³¹ Perhaps the differences are due to the quality or method of preparation. We have shown that our effects are mediated by PEDF by showing that we can neutralize the biphasic effect with anti-PEDF (see Fig. 4 ). Thus, we are confident that our data are strictly the result of the influence of PEDF. Second, we used a continuous-delivery system to administer PEDF. The ROP model treated 5- to 7-g mice intraperitoneally from postnatal day (P)7 to P12. We treated 20- to 25-g adult mice. We could not get a significant effect on CNV by intraperitoneal injection, and intravitreal injection augmented CNV (not shown). Third, the present system involves adult choroidal angiogenesis based on injury and inflammation. The ROP model involves neonatal retinal angiogenesis induced by hyperoxia and hypoxia. Perhaps the difference lies in the endothelial cell types involved in the model. 

The augmenting effects of PEDF were obtained at high concentrations that appeared to exceed the physiological levels. However, that there are dual effects on angiogenesis suggests that caution should be exercised before this agent is administered to patients. Because augmentation of CNV has not been reported when the protein is delivered in a viral vector, ^{13 15} perhaps this indicates that this a safer method. However, because the clinical trial is a dose-escalation study, ¹⁶ careful studies should be performed to determine the efficacy of PEDF over a wide range of concentrations. In addition, further studies into the mechanism of action of PEDF may provide insight into the use of this important compound as a therapeutic agent. 

 Supported by National Eye Institute Grants EY12826, EY06765, and EY02687; ORTC Grant (NIH 5 P60 DK20579) Research to Prevent Blindness; and the Foundation Fighting Blindness.

 Submitted for publication February 18, 2004; revised April 20, July 30, and August 12, 2004; accepted September 3, 2004.

 Disclosure: R.S. Apte, None; R.A. Barreiro, None; E. Duh, None; O. Volpert, None; T.A. Ferguson, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Effect of PEDF on choroidal neovascularization. A krypton laser was used to induce choroidal neovascularization (CNV) in the eyes of C57/BL6-wild type mice. After 7 days, the animals were perfused with 3% FITC-conjugated high-molecular-weight dextran (2000 kDa). After enucleation, sclerochoroidal wholemounts were created. Bruch’s membrane was stained with a Cy-3–conjugated anti-elastin antibody. CNV was identified as FITC-perfused vessels above the plane of Bruch’s membrane on confocal microscopy. PEDF was administered by the use of a miniosmotic pump that holds a total volume of 200 μL. Indicated concentrations of PEDF were 90, 180, and 360 μg/mL over the treatment period. Values are reported in cubic micrometers, representing the volume of the neovascular complex ± SE *Significantly different from the control based on Student’s t-test.

Figure 2.

 

Confocal images of neovascular complexes. Representative images of the neovascular complexes in untreated control mice (A) and mice treated with 90 μg/mL PEDF (B), 180 μg/mL PEDF (C), and 360 μg/mL PEDF (D).

Figure 3.

 

Effect of anti-PEDF antibody on CNV. A krypton laser was used to induce choroidal neovascularization (CNV) in the eyes of C57/BL6-wild-type mice. PEDF (360 μg/mL) was administered by the use of a subcutaneous miniosmotic pump that holds a total volume of 200 μL. A second group of mice was given neutralizing anti-PEDF (360 μg/mL) antibody in the same pump, along with the high-dose PEDF preparation. Another group was given anti-PEDF (360 μg/mL) antibody in a separate pump from that used for the high-dose PEDF preparation. CNV was quantified at day 7. Values are reported in cubic micrometers, representing the volume of the neovascular complex ± SE. *Significantly different from control based on Student’s t-test.

Figure 4.

 

Effect of PEDF on endothelial cell migration. Increasing concentrations of PEDF were added, alone or in combination with 10 ng/mL bFGF. Data represent, as a percentage, control migration compared with migration in the presence of 10 ng/mL bFGF.

Figure 5.

 

Vascular tube formation of HMVECs on basement membrane. Cells were left untreated (A), or treated with 5 μg/mL (B), 25 μg/mL (C), or 50 μg/mL (D) PEDF. Micrographs were taken 8 hours later. Magnification, ×100.

Figure 6.

 

VEGF production from vascular endothelial cells. HMVECs were grown in EBM with 10% FCS. They were plated at 5 × 10⁵ cells/mL in serum-free medium overnight. Indicated concentrations of PEDF were then added, and the supernatants were harvested 24 hours later. VEGF concentration was determined by ELISA. Values are expressed in picograms of VEGF per milligram total protein.

Figure 7.

 

Effect of anti-VEGF antibody on PEDF-mediated CNV. A krypton laser was used to induce choroidal neovascularization (CNV) in the eyes of C57/BL6 wild-type mice. PEDF (360 μg/mL) was administered by the use of a subcutaneous miniosmotic pump. Other groups were given 360 μg/mL of neutralizing anti-VEGF(1) or 720 μg/mL anti-VGFF(2) antibody in a separate pump, with or without PEDF. Control animals received 360 μg/mL of BSA via a pump. CNV was quantified at day 7. Values are reported in cubic micrometers, representing the volume of the neovascular complex ± SE. *Significantly different from the BSA-treated control, based on Student’s t-test.