Evidence for Pigment Epithelium–Derived Factor Receptors in the Neural Retina

Abstract

purpose. The neurotrophic activity of pigment epithelium–derived factor (PEDF), an extracellular factor present in the retina, is mediated by binding to cell-surface receptors in responsive cell cultures. In the present study, the expression of PEDF receptors in native neural retinas from adult steers was examined.

methods. Binding reactions were performed with ¹²⁵I-PEDF and fluoresceinated PEDF using plasma membranes, detergent-soluble membrane proteins, or cryosections of retina from adult bovine eyes. Radioligand-binding and competition analyses were performed with a computer-assisted program. Ligand blot analysis of detergent-soluble membrane proteins was performed with ¹²⁵I-PEDF followed by autoradiography. Ligand-affinity column chromatography of detergent-soluble membrane proteins was performed with PEDF-coupled resin followed by SDS-PAGE. Binding of fluoresceinated PEDF to retina cryosections was detected by confocal microscopy.

results. Radioligand-binding assays showed that ¹²⁵I-PEDF bound in a specific and saturable fashion to one class of sites on retina membranes (K _d = 2.5–6.5 nM; maximum binding [B _max] = 1–48 × 10¹⁰ sites/retina). A peptide of 44 amino acids (44-mer), identified as the receptor-binding region of PEDF, competed efficiently for ¹²⁵I-PEDF binding to retina membranes with kinetics similar to the full-length PEDF. Ligand blot analysis and ligand–affinity chromatography revealed a specific and high-affinity PEDF-binding protein of ∼85 kDa in retina plasma membranes. Confocal microscopy showed that fluorescein-conjugated PEDF stained exclusively the inner segments of photoreceptors and cells of the ganglion cell layer in retinal cryosections.

conclusions. Altogether, these data conclusively demonstrate the existence of PEDF receptors discretely distributed on the surface of cells from the adult neural retina of bovine eyes. Furthermore, they provide evidence for the direct action of PEDF on photoreceptor and ganglion cell neurons and an anatomic basis for studies to assess PEDF neurotrophic effects on the adult retina.

The development, morphogenesis, and survival of the neural retina rely on growth, trophic, and survival factors derived mostly from the adjacent retinal pigment epithelium (RPE). Pigment epithelium–derived factor (PEDF) was first identified as an RPE-derived protein with potent neuronal differentiating activity in human retinoblastoma cells. ¹ Recently, it was reported that PEDF protects rat retinal neurons from hydrogen peroxide–induced cell death in culture ² ; transiently delays the death of photoreceptor cells in mouse models of retinitis pigmentosa, retinal degeneration (rd/rd), and retinal degeneration slow (rds/rds) mice ³ ; protects rat photoreceptors cells from light damage ⁴ ; and has a morphogenetic effect on photoreceptor neurons of Xenopus laevis. ⁵ In addition to its effects on retina cells, PEDF has neuronal survival and differentiating activities in primary cultures of rat cerebellar granule cell neurons, ^{6 7 8} developing primary rat hippocampal neurons, ⁹ and avian and murine spinal motor neurons. ^{10 11} Thus, PEDF is a multipotent neurotrophic factor that may play a neuroprotective role in the retina in vivo and could be used as a therapeutic agent for the treatment of retinal degeneration. 

PEDF is a natural extracellular component of the retina. ^{12 13 14 15} The localization of PEDF mRNA by in situ hybridization and Northern blot analysis reveals a distribution of high levels in the RPE and undetectable levels in the neural retina of bovine and mouse eyes. ^{15 16 17} Immunoblot and immunohistochemistry analyses show that the PEDF protein localizes to the interphotoreceptor matrix in bovine, human, mouse, and rat eyes, and that cells of the RPE can secrete soluble PEDF into the media. ^{1 12 14 18 19} These observations reveal that, in the native eye, the RPE is the main source of PEDF, which produces it as a diffusible factor toward the neural retina. 

PEDF is a glycoprotein (50,000 molecular weight) that belongs to the superfamily of serine protease inhibitors (serpin) by sequence homology. ^{16 18 20 21} However, it has no inhibitory effect on serine proteases. ^{22 23 24 25} Structure–function studies have shown that a region toward the amino end of the polypeptide (44-mer, amino acid positions 78-121 of the human PEDF) confers neurotrophic activity on the PEDF polypeptide, whereas its homologous serpin-reactive site, near the carboxy end, is not essential for biologic activity. ^{8 23 25 26} PEDF has high binding affinity for cell-surface receptors in human retinoblastoma Y-79 cells and rat cerebellar granule cell neurons (K _d ≈ 3 nM), identified as plasma membrane proteins of approximately 80 kDa. ²⁶ In addition, the binding of PEDF to these receptors was competed efficiently by the 44-mer peptide. Thus, the mechanism of action for the neurotrophic activity of PEDF is independent of protease inhibition but mediated by interactions between a region contained in the 44-mer and cell-surface receptors expressed in target cells. 

Given these biological activities of PEDF and its localization adjacent to the retina, investigation of the expression of PEDF receptors in the neural retina may yield important insights into the physiological relevance of PEDF–receptor interactions and distribution of PEDF’s activity among target cells in the retina. Because the binding of PEDF to its receptor is the first step mediating its biological effects, ligand-binding assays were developed for the bovine retina. We have used recombinant human PEDF ²⁴ labeled with ¹²⁵I or fluorescein as the ligand for membrane extracts or cryosections of bovine retina, as well as unlabeled PEDF and the 44-mer peptide to further characterize the specificity of the binding. The present data provide evidence for spatially distributed neurotrophic receptors for PEDF in the neural retina. 

Methods

PEDF and PEDF-Derived Peptides

Recombinant human (rh)PEDF protein was purified from the conditioned media of BHK cells containing the expression vector pMA-PEDF with a full-length human PEDF cDNA. ²⁴ Synthetic 34-mer and 44-mer peptides were designed from amino acid positions 44-77 (DPFFKVPVNKLAAAVSNFGYDLYRVRSSMSPTTN) and 78-121 (VLLSPLSVATALSALSLGAEQRTESIIHRALYYDLISSPDIHGT) of the human PEDF sequence. ²⁶ rhPEDF was labeled with ¹²⁵I, using chloramine-T or iodination reagent (Iodobeads; Pierce, Rockford, IL). ¹²⁵I-PEDF has a specific activity of 0.25 to 2 × 10⁸ disintegrations per minute (dpm)/μg, a concentration of 2 to 6 × 10⁶ dpm/μl, and a trichloroacetic acid precipitable radioactivity of 82% to 94%. ¹²⁵I-PEDF has the structural, immunologic, and biologic characteristics of the unlabeled protein and is a specific ligand for PEDF receptors. ²⁶ rhPEDF was chemically coupled with fluorescein-5-EX succinimidyl ester (Molecular Probes, Eugene, OR), according to the manufacturer’s instructions. Briefly, a mixture of 10 μM rhPEDF and 150 μM fluorescein in 50 mM NaHCO₃ (pH 8.5; final volume of 1 ml) was incubated at 25°C for 2 hours. The reaction was quenched by adding 100 μl of 1.5 M monoethanolamine (pH 8.5) and incubating at 25°C for 2 hours. The protein was separated from free fluorescein by ultrafiltration with a microconcentrator (Centricon-30; Millipore Corp., Bedford, MA) and washed with 8 ml phosphate-buffered saline (PBS; 9 g/l NaCl, 0.144 g/l KH₂PO₄, 0.795 g/l Na₂HPO₄ [pH 7.4]). The final fluoresceinated PEDF (Fl-PEDF) sample, contained an average of two fluorescein molecules per PEDF molecule. 

Membrane Extracts from Bovine Retina

All preparation procedures were performed at 4°C. Fresh adult bovine eyes (J. W. Trueth & Sons, Baltimore, MD) were dissected below the iris, the vitreous removed from the inner retinal surface, and the neural retinas gently separated from the pigment epithelium with forceps. Retinas were homogenized in a solution of cold 0.32 M sucrose in Tris-buffered saline (TBS; 20 mM Tris/HCl [pH 7.5] and 150 mM NaCl) containing protease inhibitors (1 mM aminoethyl-benzenesulfonyl fluoride hydrochloride [AEBSF], 5 μg/ml aprotinin, 1 μg/ml pepstatin, and 0.5 μg/ml leupeptin) at 7.5 ml per retina with a homogenizer (Polytron model 3000; Brinkman Instruments, Westbury, NY) set at 10,000 rpm for 20 seconds. The homogenized material was separated from tissue and cellular debris by centrifugation at 1000g for 10 minutes and was subjected to ultracentrifugation at 80,000g for 30 minutes. The pellets, enriched in membranes were resuspended in cold 1% bovine serum albumin (BSA) in PBS (1.5 ml per retina) and constituted the retinal membrane extracts. 

Solubilization of membrane proteins was performed as described previously. ^{26 27} Briefly, membrane fractions were prepared as just described except that the homogenization buffer was 20 mM HEPES (pH 7), 100 mM KCl, 1 mM EDTA with protease inhibitors at 10 ml per retina. The membrane pellets were resuspended gently in buffer D (20 mM sodium phosphate [pH 6.5], 150 mM NaCl, 10% glycerol, 1 mM CaCl₂, and 0.5% 3-([3-cholamidopropyl]dimethylammonio)-2-hydroxy-1-propanesulfonate[ CHAPS]) at 0.4 ml per retina and centrifuged at 175,000g, 30 minutes at 4°C. The supernatant was transferred to a new tube and the pellet extracted a second time by the same procedure. The supernatants from both extractions were pooled (a final volume of approximately 0.8 ml per retina) and constituted the detergent-soluble membrane protein fraction, which was used immediately or stored at− 20°C until further use. 

¹²⁵I-PEDF Binding Assays

The method of choice has been widely used and has a mechanism of retention of receptors on polyethylenimine-treated glass fiber filters based mainly on ionic interactions. ²⁷ Integral membrane proteins tend to be acidic. Polyethylenimine binds strongly to glass, which is negatively charged, and the resultant polycationic polyethylenimine-coated glass should bind polyanions strongly. Because binding of receptors to polyethylenimine filters is rather insensitive to ionic strength, the ionic phenomenon is thought to be supplemented by hydrophobic forces and hydrogen binding. Binding reactions were performed by adding ¹²⁵I-PEDF to membrane extract suspension (0.15 ml) or detergent-soluble membrane fraction (0.08–0.15 ml) and incubations at 4°C. The specific ¹²⁵I-PEDF binding to membrane extracts did not change significantly with incubations between 1.5 and 16 hours. The reaction was stopped by adding 5 ml cold 1% BSA-PBS and immediately separating free and bound ligand by filtration under vacuum through GF/C glass fiber filters (Whatman International Ltd., UK) presoaked in 0.3% polyethylenimine. The filters were washed twice with cold 1% BSA-PBS, allowed to dry, and placed in vials with 5 ml scintillation cocktail. The radioactivity in the filters was determined with a scintillation β-counter (model LS3801; Beckman, Fullerton, CA). Nonspecific binding was defined as bound radioactivity in the presence of a 40-fold molar excess of rhPEDF over the radioligand and specific binding as bound radioactivity minus nonspecific binding. Each data point corresponds to the average of triplicate assays. Data were analyzed (Prism, ver. 3; GraphPad, San Diego, CA) for nonlinear regression with one-site binding and competition equations. 

Ligand Blot

Detergent-soluble membrane proteins were resolved by SDS-PAGE under nonreducing conditions and transferred to a 0.2-μm nitrocellulose membrane. The membrane was first washed with 1% NP-40 in TBS for 15 minutes and then twice with TBS at 25°C for 10 minutes each. The blot was incubated with blocking solution (1% BSA in TBST, containing TBS with 0.05% Tween-20) at 25°C for 2 hours and then with 2 nM ¹²⁵I-PEDF in blocking solution at 4°C for 16 hours. The blot was washed three times with TBST at 25°C for 15 minutes to remove the unbound ligand, air dried, and exposed to x-ray film (BioMax ML; Eastman Kodak Co., Rochester, NY) to detect bound radioligand by autoradiography. 

Ligand-Affinity Column Chromatography

Fresh detergent-soluble membrane fractions from retina were subjected to PEDF–affinity chromatography by a method described previously. ²⁶ Briefly, rhPEDF was coupled to beads of preactivated hydrophilic, cross-linked bis-acrylamide/azlactone copolymers (3M Emphaze Ultralink; Pierce, Rockford, IL). Detergent-soluble proteins obtained from 10 bovine eyes were passed through a column of resin without ligand (1.8 ml). The unbound material was mixed with PEDF-coupled resin (1.8 ml; 6 mg PEDF/ml resin) and gently rotated at 4°C for 1 hour. The material was packed in a column, washed with buffer D (20 column volumes or until absorbance at 280 nm was undetectable), followed by 1 M NaCl in buffer D (10 column volumes). The bound material was eluted with 0.1 M glycine buffer (pH 11) 10% glycerol, 1 mM CaCl₂, 0.15 NaCl, and 0.25% CHAPS (10 column volumes). Eluted proteins were concentrated to 100 μl by ultrafiltration with microconcentrators (Centricon-30; Millipore). The microconcentrators were washed twice with 100 μl buffer D. 

Heparin-Affinity Column Chromatography

Heparin-affinity column chromatography was performed as described previously. ²⁸ Briefly, rhPEDF or Fl-PEDF (30μ g) was incubated with heparin immobilized on acrylic beads at 4°C for 30 minutes. Unbound material (flow-through) was removed by extensive washes with binding buffer. Bound proteins were eluted with a step gradient of 0.3 M and then 1 M NaCl in buffer H (20 mM sodium phosphate [pH 6.5], 20 mM NaCl, and 10% glycerol). Fractions were concentrated by ultrafiltration with microconcentrators. 

Fl-PEDF Binding to Retinoblastoma Y-79 Cells

Human retinoblastoma Y-79 cells (6 × 10⁵ cells/ml) were incubated in serum-free medium at 37°C for 16 hours. ²² Binding started by adding Fl-PEDF (20 nM) and BSA (0.1%) to the culture medium and incubating at 4°C for 90 minutes. Unbound ligand was removed from the cells by centrifugation and three washes with 1% BSA-PBS. The cells were mounted in antifading solution (SlowFade; Molecular Probes) and scanned with coherent light of 488 nm for fluorescein visualization under a laser scanning microscope (model 510; Carl Zeiss, Oberkochen, Germany). 

Fl-PEDF Binding to Cryosections of Bovine Retina

Bovine eyes were sectioned below the iris, the vitreous was removed, and a section of the posterior part of the eye encompassing the retina, choroid, and sclera, was excised (0.5 × 2.5 cm²) and embedded in optimal temperature cutting (OCT) compound (Sakura Finetek USA, Inc., Torrance, CA). Cryosections (7 μm thick) were prepared on glass slides and used for in situ Fl-PEDF binding. The frozen sections were prewashed in ice-cold 1 M NaCl in 1% BSA-PBS and then in 1% BSA-PBS for 5 minutes each. The prewash with 1 M NaCl did not affect the binding of Fl-PEDF and served to preserve the cryosection from disintegrating through subsequent incubation reactions. The cryosections were then incubated with Fl-PEDF in 1% BSA-PBS plus 1 protease inhibitor cocktail tablet per 50 ml (Complete; Roche, Indianapolis, IN), in a humid chamber at 4°C for 30 minutes, washed with ice cold 1% BSA-PBS (5 minutes) to remove unbound ligand, and fixed with 10% formalin (3 minutes). Sections were incubated with 1 μM 4′,6-diamidino-2-phenylindole (DAPI; Molecular Probes) at 25°C for 5 minutes, washed with PBS, and mounted with antifading solution (SlowFade; Molecular Probes). Sections were scanned with coherent light of 366 nm (for visualization of DAPI staining), 488 nm (fluorescein), and 546 nm (background in the red channel), using the laser scanning microscope (Zeiss). Confocal images were obtained, maintaining the same microscope settings for all samples to allow comparisons between the treated sections and the control samples. 

Western Blot Analysis

Western transfers and immunoreactions were performed as described before. ¹² Briefly, immunoreactions with rabbit polyclonal antiserum to human PEDF (Ab-rPEDF; diluted 1:1000) ¹² or anti-Na⁺,K⁺-ATPase (0.5μ g/ml; Upstate Biotechnology, Lake Placid, NY) were followed by sequential incubations with biotinylated anti-rabbit IgG (1:1000) and ABC complex (Vector Laboratories, Inc., Burlingame, CA) and immunostaining with 4-chloro-naphthol. Immunoreactions with a monoclonal antibody for the human mitochondrial membrane protein of the oxidative complex IV, cytochrome oxidase subunit I, was with anti-COX-I (1D6-E1-A8; Molecular Probes) at 2 μg/ml in PBST (0.05% Tween-20 in PBS) at 4°C for 1 hour. This was followed sequentially by washes with PBST, incubation with horseradish peroxidase-conjugated goat anti-mouse IgG (diluted 1:1000, Roche), washes with PBST, incubation in 10 ml of chemiluminescent substrate solution (LumiLight; Roche), and exposure to x-ray film (Biomax ML; Eastman Kodak, Co.) with development to visualize the immunoreaction. 

Other Assays

SDS-PAGE was performed with 10% to 20% or 4% to 12% polyacrylamide gradient gels in SDS-tricine or SDS-Tris-glycine, respectively (Novex, San Diego, CA). Protein concentration was determined using a protein assay (Bio-Rad Laboratories, Hercules, CA). 

Results

¹²⁵I-PEDF Binding to the Neural Retina

To examine the binding of PEDF to putative receptors in the retina, we prepared membrane extracts from bovine neural retinas and performed radioligand-binding assays with ¹²⁵I-PEDF, a specific ligand for PEDF receptors in cell cultures. ²⁶ The binding reaction mixtures were incubated at 4°C to avoid proteolytic degradation and denaturation of ligand and/or membrane proteins. The binding data were obtained with a given amount of ¹²⁵I-PEDF and increasing concentrations of unlabeled ligand. We found that specific PEDF binding increased with ligand concentrations ranging from 0 to 10 nM, in a saturable fashion (Fig. 1) . Nonlinear regression of the binding data with a one-binding-site equation revealed that PEDF bound with high affinity to 2.54 × 10¹¹ sites/retina with a K _d of 2.524 nM. Several replicate experiments performed with membrane extracts from different batches of retinas revealed similar kinetics, with K _d ranging between 3.1 and 6.5 nM and maximum binding (B _max) = 0.1 to 4.8 × 10¹¹ sites/retina (Table 1) . The data did not converge with a two-binding-site equation. The K _d for the retina receptors is in the same order of magnitude as the that for receptors on cultured retinoblastoma and cerebellar granule cells. ²⁶  

Competition of ¹²⁵I-PEDF Binding with 34-mer and 44-mer Synthetic Peptides

The region of the 44-mer peptide represents the neurotrophic receptor binding domain of PEDF. To determine how the 44-mer affects PEDF binding to retina receptors, radioligand competition assays were performed with ¹²⁵I-PEDF and the synthetic 44-mer peptide in the concentration range of 0 to 5 nM. Figure 2A shows that increasing concentrations of 44-mer competed for the binding of 0.25 nM ¹²⁵I-PEDF with kinetics similar to rhPEDF. Nonlinear regression of the data with an equation for one-site competitive binding revealed median effective concentrations (EC₅₀) between 4.4 and 13.7 nM PEDF and 2.05 and 5.96 nM 44-mer, in three different batches of retina. The data did not converge with a two-site competitive binding equation. The negative control samples, 34-mer peptide (hPEDF positions 44-77) and cytochrome c (Pharmacia, Piscataway, NJ), did not have a significant effect on binding (Fig. 2B) . These results demonstrate that the 44-mer peptide behaved as a competitive inhibitor of PEDF binding to receptors in the retina. 

PEDF Binding to Solubilized Plasma Membranes

To examine the biochemical characteristics of the PEDF-binding components in the neural retina, we prepared detergent-soluble extracts from membranes of bovine neural retinas. Western analyses with plasma and mitochondrial membrane markers revealed that the detergent membrane extracts were highly enriched in plasma membranes (Fig. 3A) . Radioligand-binding assays demonstrated that, after solubilization with CHAPS, the membrane preparations retained as high as one-tenth of the ¹²⁵I-PEDF-binding activity (data not shown). Losses in activity may have been due to inefficient solubilization, improper protein folding, or instability in detergent of the PEDF-binding components. Ligand blot analysis performed with detergent-soluble membranes and 2 nM ¹²⁵I-PEDF identified a component with high binding affinity for ¹²⁵I-PEDF and an electrophoretic migration of an 85-kDa protein identical with that in retinoblastoma cells (Fig. 3B) . The binding of ¹²⁵I-PEDF to the 85-kDa protein band was found consistently among three different bovine neural retina preparations as opposed to other bands of slower and faster migration patterns. Competition with an excess of rhPEDF decreased the signal for ¹²⁵I-PEDF binding of the 85-kDa protein. A computer program (NIH Image software; Scion Corp., Frederick, MD) was used to compare the relative pixel density of the bands, with and without excess unlabeled ligand. The background was adjusted to give similar density around the 85-kDa protein band. The density of the 85-kDa protein band with excess of unlabeled ligand (Fig. 3B , lane 6) was 50% to 60% of that without unlabeled ligand (lane 5). Preincubations of detergent-soluble fractions at 37°C resulted in the disappearance of such activity (data not shown), implying that the 85-kDa PEDF-binding activity was heat-inactivated. PEDF-affinity column chromatography of the detergent-soluble fraction revealed a protein of approximately 85-kDa with binding affinity for immobilized PEDF (Figs. 3C 3D) , in agreement with the ligand blot results. Other minor bands, migrating as 60- to 66-kDa proteins, were not observed in a second experiment performed with a different batch of retinas and PEDF-affinity resins. These data reveal a specific PEDF-binding protein of ∼85 kDa in retina plasma membranes consistent with the presence of a cell-surface receptor protein in the bovine neural retina. 

Fl-PEDF as a Ligand for PEDF Receptors

Before studying the distribution of PEDF binding in the retina, we investigated structural and binding properties of the ligand of choice. Fluorescein-5-EX succinimidyl ester was chemically coupled to primary amines of lysines and/or the amino-end group of rhPEDF. The chemically modified PEDF migrated as a 50-kDa protein and immunoreacted with Ab-rPEDF, similar to the unmodified PEDF (see Fig. 4D ). To determine the receptor-binding activity of Fl-PEDF, we used human retinoblastoma Y-79 cells because they contain PEDF cell-surface receptors. ²⁶ Cells incubated with Fl-PEDF exhibited fluorescein-staining on their surface, which, as visualized by confocal microscopy, was more intense in a few cells of each aggregate (Fig. 4A) , whereas those incubated with Fl-PEDF and an excess of rhPEDF showed a significant decrease in staining (Fig. 4B) , demonstrating a specific and competitive Fl-PEDF binding to Y-79 cells. 

PEDF has binding affinity for glycosaminoglycans, which is mediated by ionic interactions between these polyanions and a positively charged region on the surface of the PEDF protein. ²⁸ To determine how fluorescein coupling affects the glycosaminoglycan binding, we subjected Fl-PEDF to heparin-affinity column chromatography. Fl-PEDF was not retained by the heparin-affinity resin as opposed to the unmodified PEDF, which required an increase of NaCl concentration to be eluted from the column (Fig. 4D) . Together, these results indicate that Fl-PEDF retained its ability for interacting with cell-surface receptors, but lost its ability to bind glycosaminoglycans. 

Localization of PEDF Binding Sites in the Bovine Retina

Fl-PEDF was used to determine the distribution of PEDF binding sites in cryosections of bovine retina. Confocal microscopy of cryosections incubated with Fl-PEDF showed dense staining in the region of the inner segments (IS) of photoreceptor cells (Figs. 5A 5C 5G) and in the ganglion cell layer (GCL; Figs. 5A 5J ). The fluorescein signal decreased significantly when cryosections were incubated with Fl-PEDF plus an excess of unmodified protein (Fig. 5H) , in the absence of ligand (Figs. 5D 5F) or with anti-rabbit IgG labeled with fluorescein (data not shown), as a negative control, indicating a specific and competitive Fl-PEDF binding to the retina. The binding pattern to the IS was found consistently among four different bovine eyes. In three of the specimens, the staining was also observed in large ganglion cells, but not all the cells were stained (Fig. 5L) . The RPE appeared to have intrinsic fluorescence, as illustrated by comparing Figure 5I with Figures 5G and 5H . Detection of Fl-PEDF on cryosections by amplification of fluorescein signal by colorimetric staining of anti-fluorescein and visualization by light microscopy produced identical results. ²⁹ These results demonstrate a discrete spatial distribution of specific Fl-PEDF–binding sites in the photoreceptors and cells of the GCL of the bovine neural retina. 

Discussion

We have demonstrated that the bovine neural retina contains cell-surface receptors for PEDF. The data reveal high-affinity sites in the retina that have PEDF-binding characteristics similar to receptors on the surface of cells that respond to neurotrophic stimulation by PEDF. ²⁶ Several lines of evidence support these conclusions. First, ¹²⁵I-PEDF and Fl-PEDF bound specifically to membranes and cryosections of bovine neural retina, respectively. Second, the K _d for PEDF binding to receptors in the neural retina (2.5–5.5 nM), retinoblastoma cells (2.63 ± 0.93 nM), and cerebellar granule cells (3.18 ± 0.93 nM) were within the same order of magnitude. Third, the full-length PEDF and 44-mer peptide were equal competitors for ¹²⁵I-PEDF binding to the high-affinity receptors on both the neural retina and retinoblastoma cells. In addition, the binding of ¹²⁵I-PEDF to cell membranes from neural retinas and retinoblastoma cells was inhibited by Ab-rPEDF, a blocking antiserum of neurotrophic activity, and not by the serpin ovalbumin. ²⁹ Fourth, a specific PEDF-binding protein among plasma membranes of retina, retinoblastoma, and cerebellar granule cells had comparable biochemical characteristics. 

These similarities suggest that PEDF interacts with a cell-surface protein in the bovine neural retina that is homologous to receptors in human retinoblastoma and rat cerebellar granule cells. The fact that the 44-mer peptide, the receptor-binding site of PEDF, blocked binding of PEDF to retina plasma membranes points to interactions with neurotrophic receptors present in the surface of retina cells. This conclusion is in agreement with in vitro survival effects of PEDF on retina cell cultures and in vivo protective effects on photoreceptor cells of the rd/rd and rds/rds mice. ³  

A spatial structural model for PEDF reveals that the 44-mer region, the receptor-binding site, is located in a distinct area opposite the glycosaminoglycan-binding region. ²⁸ The 44-mer has no amino acids with primary amines and is located in an area that has a negative electrostatic potential. In contrast, the glycosaminoglycan-binding region is densely populated with lysines that confer a basic electrostatic charge to the surface of the protein and are available to interact with the negatively charged glycosaminoglycans. Whereas ¹²⁵I labeling modifies tyrosine groups, fluorescein labeling modifies primary amines, altering the basic electrostatic charge on the surface of the protein. We have shown that the chemical modification of PEDF’s primary amines with either fluorescein (Fig. 5D) or biotin ²⁸ abolishes the binding to glycosaminoglycans, but not to the PEDF receptor on the cell-surface of retinoblastoma Y-79 cells (Fig. 5A) . In addition, the Fl-PEDF is an active neurotrophic factor that shares the neuronal differentiating and survival activities in retinoblastoma and cerebellar granule cells with its unlabeled counterpart (personal observations, Vicente Notario and Joan P. Schwartz, 2000). Therefore, the use of Fl-PEDF allows the detection of interactions with the PEDF receptor while excluding those with glycosaminoglycan-rich areas of cells and tissue. These characteristics confer unique qualities on Fl-PEDF as a ligand for tissues and cells. 

The present data also reveal the localization of PEDF binding sites in the neural retina. We found that the distribution of Fl-PEDF binding predominated in the IS of photoreceptor cells and also in cells of the GCL. Previous reports indicate that retinal pigment epithelial cells express the PEDF gene and secrete the mature PEDF protein (the ligand) into the interphotoreceptor matrix next to the neural retina. ^{12 14 15 16 18 19} The Fl-PEDF–binding sites in the IS of the photoreceptor cells probably represent the cell-surface receptors available to interact with the extracellular ligand and agree with PEDF protective and morphogenetic effects on photoreceptors of the rd and rds mice, ³ rat, ⁴ and Xenopus laevis. ⁵ Thus, PEDF may act directly on photoreceptor cells through a cell-surface receptor. 

The effect of PEDF on ganglion cells is unknown, however, and the data on Fl-PEDF–binding suggest for the first time that ganglion cells may be targets for PEDF activity. Purified PEDF of 49,500 molecular weight from bovine interphotoreceptor matrix has a retention time in a TSK-300 gel column (TosaHaas, Japan) by HPLC that is slightly behind than that of ovalbumin (a serpin of 43,000 molecular weight). ¹² Because the Stokes radius of ovalbumin (hen egg) is 3.05 nm (available at http://itsa.ucsf.edu/∼hdeacon/Stokesradius.html), we estimate that the one for PEDF is not larger than 3.05 nm. The size of PEDF suggests a large protein and indicates a certain degree of difficulty in diffusing through the outer limiting membrane (OLM). Although we have not designed experiments to validate PEDF’s diffusion through the OLM in the bovine retina, the following observations suggest that naturally occurring PEDF molecules can cross the OLM in other species: (1) Biotinylated-PEDF of 50 kDa injected in the vitreous of adult mice was detected in the neural retina and RPE-choroid after injection (see Ref. ³ ). (2) The intravitreally injected recombinant human PEDF of 50 kDa had an effect on photoreceptors of rd and rds mice throughout the retina. ³ In addition, immunohistochemistry of albino rat retina with polyclonal antiserum to human PEDF shows some PEDF immunoreactivity distributed in cells of the GCL, ¹⁹ and as recently reported, PEDF gene expression is present in cells of the GCL of the human retina. ³⁰ Together, these observations point to colocalization of PEDF receptors and ligand in cells of the GCL and insinuate possible functional effects of PEDF on cells of the GCL. 

Comparison with other neurotrophic factors indicates that ciliary neurotrophic factor (CNTF) and brain-derived neurotrophic factor (BDNF) also have protective effects on photoreceptors. ^{31 32 33 34} Localization of CNTF receptors by immunohistochemistry of the avian retina reveals a distribution in the outer segments of mature rods. ³⁵ In contrast, BDNF activates tyrosine kinase (Trk)-B receptors ³⁶ distributed in retinal pigment epithelial cells and cells of the inner retina, ³⁷ but not in photoreceptor cells of mouse ³⁸ or rat. ³⁹ Thus, the similarities between PEDF’s and CNTF’s protective effects and receptor distribution suggest that, as opposed to BDNF, both factors may act by interacting directly on photoreceptor cells. 

In summary, using two PEDF ligands, we have demonstrated for the first time the presence of PEDF receptors in the neural retina. The demonstration of ¹²⁵I-PEDF and Fl-PEDF binding in the bovine retina supports a role for this factor in the adult retina and provides an anatomic basis for investigations into the in vivo activity of PEDF in the neural retina. Further studies are necessary to confirm that the binding activity described herein is directed to functional receptors that, after interactions with their ligands, trigger the signal transduction events for neurotrophic activity on the neural retina. However, our data correlate with PEDF’s effects on the survival and morphogenesis of photoreceptor cells in vivo and retina cells in culture, ^{2 3 4 5} suggesting that the PEDF binding sites on photoreceptor cells correspond to functional receptors. These observations imply that, in addition to binding in vitro, the PEDF–receptor interactions may serve to localize and direct PEDF activity in the neural retina. 

²  Present affiliation: Departamento de Neurociencias, Universidad del País Vasco, Vizcaya, Spain.

 Submitted for publication February 16, 2001; revised July 2, 2001; accepted September 5, 2001.

 Commercial relationships policy: P (SPB); N (all others).

 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.

 

Binding of ¹²⁵I-PEDF to membranes from neural retinas. Each binding assay contained membrane extracts corresponding to one-tenth of a bovine neural retina, ¹²⁵I-PEDF as the ligand, and increasing amounts of rhPEDF. The binding data were obtained with 0.25 nM ¹²⁵I-PEDF and incubations at 4°C for 16 hours. Specific binding was calculated by subtracting the nonspecific binding from the total binding and using the specific radioactivity for each point. Data analysis was performed to obtain nonlinear regression and a one-site-binding fit. The physicochemical parameters were K _d = 2.524 nM and B _max = 3.7 × 10¹⁰ sites/retina. Inset: binding data displayed in a Scatchard plot. Each data point corresponds to the average of triplicate assays. This experiment was performed four times, producing similar results (see Table 1 ).

Table 1.

 

Summary of PEDF-Binding Parameters Obtained with Different Retinal Membrane Batches

Experiment	K d	B max	Data Points (n)	PEDF Concentration Range (nM)
1	2.524 ± 0.235	42.27 ± 1.46	9	0–10
2	6.500 ± 0.393	81.10 ± 1.73	8	0–51
3	5.378 ± 0.47	61.25 ± 2.01	6	0–26
4	3.115 ± 0.258	1.78 ± 0.04	4	0–51

 Assays for binding to membrane extracts were performed using 0.25 nM ¹²⁵I-PEDF for experiment 1 and 1 nM ¹²⁵I-PEDF for experiments 2, 3, and 4 and increasing concentrations of unlabeled rhPEDF. Binding data were analyzed using nonlinear regression with a one-binding-site equation. K_d is expressed in nanomolar ± SE and B_max in femtomoles per 0.1 retina ± SE.

Figure 2.

 

The 44-mer peptide competed with ¹²⁵I-PEDF for binding sites in neural retinas. Binding of ¹²⁵I-PEDF (0.25 nM) to bovine retina membrane extracts was performed with increasing concentrations of 44-mer, rhPEDF, 34-mer, and cytochrome c at 4°C for 16 hours. Percentage of ¹²⁵I-PEDF bound was calculated from the specific binding: 100% corresponds to binding in the absence of competitor and 0% in the presence of a 10-fold molar excess of rhPEDF. (A) Binding with 44-mer (solid line and solid circles) and rhPEDF (dashed lines and open circles); (B) binding with 34-mer (solid circles) and cytochrome c (open circles). Data were analyzed to fit curves with nonlinear regression and one-site-binding competition. Each data point corresponds to the average of triplicate assays. This experiment was performed at least twice. Assays with three different batches of retina membranes produced EC₅₀ ranging from 4.42 to 8.29 nM PEDF and from 2.047 to 5.96 nM 44-mer.

Figure 3.

 

PEDF binding to detergent-soluble plasma membrane proteins. Membranes from bovine neural retinas were isolated and proteins were extracted with CHAPS. (A) Western blot analysis of subcellular fractions from retinas with antibody to Na⁺,K⁺-ATPase, a plasma membrane marker (top blot), and antibody to cytochrome oxidase subunit I (COX-I), a mitochondrial membrane marker (bottom blot). Lane 1: crude tissue homogenate; lane 2: detergent-soluble membrane fraction. (B) Ligand blot with ¹²⁵I-PEDF. Western transfers of detergent-soluble plasma membrane proteins from human retinoblastoma Y-79 cells (positive control; lane 1) and three different batches of retinas (lanes 2-5) incubated with 2 nM ¹²⁵I-PEDF and from retina extracts as in lane 5 incubated with 2 nM ¹²⁵I-PEDF plus 400-fold molar excess rhPEDF (lane 6). SDS-PAGE was performed with 4% to 12% polyacrylamide gels. Lanes 3 and 4: two different ligand blot assays with the same membrane preparation. This experiment was performed at least seven times, producing similar results. (C) PEDF-affinity column chromatography. Detergent-soluble plasma membrane proteins from bovine retina (lane 1) were passed through resin without PEDF and the flow-through mixed with PEDF-affinity resin. After the mixture was loaded on a column, the unbound material was washed with binding buffer and collected in three fractions (lanes 2-4). Bound material was eluted with glycine buffer at pH 11 and concentrated with an ultrafiltration microconcentrator (lane 5). The filters were washed twice with binding buffer and collected (lanes 6-7). Fractions were resolved by SDS-PAGE using 10% to 20% polyacrylamide gels. (D) Similar results were obtained in a second experiment with a different batch of retinal membranes in which the membrane proteins were extracted with Triton X-114 instead of CHAPS ²⁶ and the PEDF-affinity column was prepared with a recombinant PEDF fragment derived from Escherichia coli cells bearing an expression plasmid for human PEDF with positions 44 to 418 ²² . Lane 1: membrane protein extract; lane 2: flow-through; lane 3: elution with glycine buffer at pH 11; lane 4: elution with glycine buffer at pH 11 containing 1 M NaCl. Silver-stained polyacrylamide gels are shown in (C) and (D). Arrows: migration position of a main PEDF-binding protein. Migration positions of molecular weight (MW) standards are indicated.

Figure 4.

 

Binding properties of fluoresceinated-PEDF. (A–C) Binding to receptors in retinoblastoma cells. Human retinoblastoma Y-79 cells were incubated with 20 nM Fl-PEDF (A), with Fl-PEDF plus a 200-fold molar excess of rhPEDF (B), and without ligand (C) at 4°C for 90 minutes. Bound Fl-PEDF was visualized with a confocal microscope with settings kept constant for all the samples analyzed. Photographs of representative fields are shown. (D) Binding to heparin. Western blot analysis of fractions from heparin-affinity column chromatography of Fl-PEDF (lanes 1–4) and unmodified rhPEDF (lanes 5–8) using the polyclonal antiserum to human PEDF, Ab-rPEDF. ¹² Lanes 1 and 5: load; lanes 2 and 6: flow- through; lanes 3 and 7: elutions with 0.3 M NaCl; and lanes 4 and 8: elutions with 1 M NaCl. These experiments were performed at least twice with similar results. Bar, 5μ m.

Figure 5.

 

Fl-PEDF binding sites in bovine retina. (A–C) A cryosection of retina incubated with 50 nM Fl-PEDF; (D–F) a cryosection from the same eye incubated without ligand. Fluorescein detection (A, D), cell nuclei stained with DAPI (B, E), and a composite of the two stains plus scanning for the red channel (C, F). (G–H) Fluorescein detection in the IS at a higher magnification after incubations with 20 nM Fl-PEDF (G), 20 nM Fl-PEDF plus 200-fold molar excess of rhPEDF (H), and without ligand (I). (J–L) GCL at a higher magnification after incubations with 50 nM Fl-PEDF: (J) fluorescein detection, (K) GCL nuclei stained with DAPI, and (L) a composite of (J) and (K). Asterisks align with regions of specific Fl-PEDF binding. This experiment was performed at least five times with similar results. Bar, (A–F) 150 μm; (G–L) 15μ m.

The authors thank Mary Alice Crawford for the preparation of cryosections from bovine retina; and Barbara Wiggert, Susan Gentleman, Joan Schwartz, and Vicente Notario for insightful discussions and critical reading of the manuscript.