Binding of Pigment Epithelium-derived Factor (PEDF) to Retinoblastoma Cells and Cerebellar Granule Neurons

Pigment epithelium-derived factor (PEDF) has neuronal differentiation and survival activity on retinoblastoma and cerebellar granule (CG) cells. Here, we investigated the presence of PEDF receptors on retinoblastoma Y-79 and CG cells. PEDF radiolabeled with l25I remained biologically active and was used for radioligand binding analysis. The binding was saturable and specific to a single class of receptors on both cells and with similar affinities (K d = 1.7–3.6 nm, B max = 0.5–2.7 × 105 sites/Y-79 cell; and K d = 3.2 nm, B max = 1.1 × 103 sites/CG cell). A polyclonal antiserum to PEDF, previously shown to block the PEDF neurotrophic activity, prevented the125I-PEDF binding. We designed two peptides from a region previously shown to confer the neurotrophic property to human PEDF, synthetic peptides 34-mer (positions 44–77) and 44-mer (positions 78–121). Only peptide 44-mer competed for the binding to Y-79 cell receptors (EC50 = 5 nm) and exhibited neuronal differentiating activity. PEDF affinity column chromatography of membrane proteins from both cell types revealed a PEDF-binding protein of ∼80 kDa. These results are the first demonstration of a PEDF-binding protein with characteristics of a PEDF receptor and suggest that the region comprising amino acid positions 78–121 of PEDF might be involved in ligand-receptor interactions.

Pigment epithelium-derived factor (PEDF) 1 was initially identified as a protein secreted by cultured human fetal retinal pigment epithelial cells with potent neuronal differentiating activity on retinoblastoma cells (1). Addition of PEDF at nanomolar concentrations to the media of human retinoblastoma Y-79 and Weri cells induces a neuronal phenotype, which is accompanied by the expression of the neuronal markers neuronal-specific enolase and 200-kDa neurofilament (2,3). PEDF also exhibits neurotrophic activities on primary cultures of rat cerebellar granule (CG) neurons, such as neuronal survival (4) and protection against death by glutamate neurotoxicity (5) and by apoptosis (6). Recent reports indicate that it has effects on other types of neurons. It promotes the survival and differentiation of developing spinal motor neurons (7), and can protect them against glutamate neurodegeneration (8). PEDF can also protect developing primary hippocampal neurons against glutamate neurotoxicity (9). In addition, it delays the death of photoreceptors in mouse models of inherited retinal degenerations (retinitis pigmentosa) (10). Thus, PEDF is a potential neurotrophic factor of the central nervous system and retina. PEDF is also referred to as early population doubling level cDNA (EPC-1), reflecting its up-regulation during G 0 in young but not senescent cultured fibroblasts (11), which suggests a role for the protein in cell maintenance.
PEDF is a glycoprotein with a molecular weight of 50,000 identified extracellularly in the vertebrate eye (12)(13)(14)(15). It is associated by ionic interactions with glycosaminoglycans in the interphotoreceptor matrix (16) and can be readily purified from the interphotoreceptor matrix and vitreous (12,13). We have previously shown that a polyclonal antiserum to human PEDF blocks the neuronal differentiating activity of the purified protein and of the interphotoreceptor matrix extracts on Y-79 cells (12), indicating that PEDF is the sole component of the interphotoreceptor matrix with such activity. The antiserum to PEDF also blocks the neuronal survival effects of PEDF on CG cells (5) and its antiproliferative effects on microglia (17).
The sequence of human, bovine, and mouse PEDF cDNA reveals that PEDF is a member of the serpin superfamily of serine protease inhibitors (2,11,18,19). Previous studies have suggested important roles for certain serine proteases and serpins in the development and/or pathology of the nervous system, e.g. thrombin, urokinase, and their inhibitors protease nexin-1 and plasminogen activator inhibitor (20 -23). Given these studies, we had previously attempted to determine the inhibitory capacity of PEDF against serine proteases. First, an inhibitory activity against proteases, among others thrombin, could not be demonstrated for PEDF (24,25). Second, PEDF, like the noninhibitory serpins ovalbumin, angiotensinogen, and maspin, lacks the serpin S 3 R conformational change upon cleavage of its serpin-exposed loop (25,26). Third, structure-function studies demonstrated that the serpin reactive loop located toward the carboxyl end of the polypeptide is dispensable, whereas a region toward the amino end (BA, amino acid positions 44 -121) confers the neurotrophic activity to the PEDF polypeptide (26). Altogether, these observations indicate that the mechanism of action for the neurotrophic activity of PEDF is independent of protease inhibition.
Therefore, it is of interest to investigate the binding properties of PEDF to cells for mechanistic studies. Because the binding of PEDF to its receptor is presumably the first step in the mediation of its physiological effects, basic physicochemical parameters of such binding were established. We have used cultures of human retinoblastoma cells and rat CG cells because they respond to PEDF stimuli, and we used a biologically active form of 125 I-PEDF to define the basic physicochemical parameters of such binding. We have used PEDF purified from bovine eyes (12,13), recombinant human PEDF (26), synthetic peptides derived from BA, and a polyclonal antiserum to PEDF, Ab-rPEDF (12), to further investigate the specificity of the binding. Finally, PEDF affinity column chromatography was used to isolate a PEDF-binding protein from retinoblastoma and CG cell membranes. We describe here that PEDF exhibits a saturable and specific binding to target cells for neurotrophic activity and demonstrate for the first time a PEDF-binding protein with characteristics of a PEDF receptor.
Preparation of PEDF Proteins-Native PEDF was purified from the interphotoreceptor matrix or vitreous of fresh bovine eyes, as described previously (12,13). Recombinant PEDF was synthesized by baby hamster kidney (BHK) cells containing the expression vector pMA-PEDF with a full-length human PEDF cDNA, and the recombinant PEDF protein was purified from the conditioned media (26). Synthetic peptides 34-mer and 44-mer were designed from amino acid positions 44 -77 (DPFFKVPVNKLAAAVSNFGYDLYRVRSSMSPTTN) and 78 -121 (VLLSPLSVATALSALSLGAEQRTESIIHRALYYDLISSPDIHGT) of the human PEDF sequence (GenBank TM accession number U29953), respectively, and prepared by Biosynthesis, Inc., followed by high pressure liquid chromatography purification (Ͼ90% purity) and aminoterminal sequence determination. The resulting peptides were soluble in aqueous solutions. rPEDF-BH is the recombinant human PEDF (positions 44 -121) prepared from Escherichia coli, as described before (24). Radiolabeled PEDF was prepared from either the native bovine or recombinant human PEDF protein with 125 I using an immobilized form of N-chlorobenzenesulfonamide (sodium salt), the Iodobeads TM iodination reagent (Pierce) by Lofstrand Laboratories and analyzed as described previously (16). The 125 I-PEDF sample contained Ͼ90% trichloroacetic-precipitable counts, with a specific activity ranging between 6 -7 ϫ 10 7 dpm/g and a concentration of 6 -7 ϫ 10 5 dpm/g.
Cell Cultures-Human retinoblastoma Y-79 and Weri cells were cultured in suspension in minimum Eagle's medium supplemented with 15% fetal bovine serum, 2 mM L-glutamine, and antibiotics (100 units/ml penicillin and 100 g/ml streptomycin) at 37°C in a humidified incubator under 5% CO 2 until the appropriate number of cells was available. Cells were collected by centrifugation, washed twice, resuspended in PBS, and then counted. For binding assays and membrane protein isolation, cells were cultured overnight at 1.25 ϫ 10 5 cells/ml in serum-free medium consisting of minimum Eagle's medium supplemented with 1 mM sodium pyruvate, 10 mM HEPES, 1ϫ nonessential amino acids, 1 mM L-glutamine, and antibiotics as above. CG cells were prepared from 8-day old Harlan Sprague-Dawley rat pups and cultured in monolayers in chemically defined media, as described previously (4). BHK, COS-7, and NIH3T3 were cultured in cell monolayers in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and antibiotics as above.
Neurite Outgrowth Assays-Human Y-79 retinoblastoma cells were cultured and treated for neurite outgrowth as described before (24). Briefly, 2 ml of Y-79 cell culture (1.25 ϫ 10 5 cells/ml) in serum-free medium as above plus 0.1% ITS were treated with 25 l of a solution containing PEDF protein in phosphate-buffered saline with 1% BSA. After 7 days of treatment, the cells were attached to poly-D-lysinecoated plates. The differentiation state of the cells was monitored by light microscopy at intervals after attachment.
Radiolabeled 125 I-PEDF Binding Assays-Human retinoblastoma Y-79 and Weri cell cultures (6 ϫ l0 5 cells/ml) were incubated at 4°C for 15 min before the addition of 125 I-PEDF at concentrations as indicated.
The binding reaction mixture was incubated at 4°C for 90 min, unless otherwise indicated. The reaction was terminated by the addition of 10 ml of ice-cold PBS supplemented with 0.1% BSA, immediately followed by filtration under vacuum through Whatman GF/C filters presoaked in 0.3% polyethylenimine. Finally, the filters were washed with 10 ml of the ice-cold 1% BSA in PBS. BHK, COS-7, NIH3T3, and CG cells were cultured in 24-well plates, the first three to 90% confluency and the latter containing 5 ϫ 10 5 cells/well. Cells were washed with 0.5 ml of 0.1% BSA in culturing medium (binding buffer) three times before the addition of 125 I-PEDF in binding buffer. After incubation at 4°C for 90 min, unless indicated, the unbound PEDF was washed with 0.5 ml of binding buffer three times, and then the cells were lysed by incubation with 0.5 ml of 1 M NaOH at room temperature for 30 min. Filters and cell lysates were placed in scintillation vials, mixed with 5 ml of Bio-Safe II liquid scintillation solution (Research Products International Corp.), incubated at room temperature overnight, and then mixed by extensive vortexing before determining the radioactivity using a ␤-counter (Beckman, model LS3801). Alternatively, bound and free radioligand were separated by centrifugation of cell suspensions followed by three washes with 1% BSA in PBS, and the bound radioactivity was determined in the cell pellets using a gamma counter (Wallac) or after SDS-PAGE and autoradiography. Nonspecific binding was defined as the amount of bound radioactivity in the presence of saturating concentrations of unlabeled ligand, unless otherwise indicated, and specific binding as bound radioactivity minus nonspecific binding. Data were analyzed using the Minitab statistical program and Microsoft Excel 97 for linear regression, as well as GraphPad Prism, version 3.0, for nonlinear regression and Scatchard analyses. Preparation of Detergent-soluble Membrane Extracts-Human retinoblastoma Y-79 cells (25 ϫ 10 6 ) were collected by centrifugation, washed twice with PBS, and resuspended in 100 ml of homogenization buffer (20 mM HEPES, pH 7.0, 100 mM KCl containing 1 mM EDTA, 5 g/ml aprotinin, 1 mM AEBSF, 1 g/ml pepstatin, and 0.5 g/ml leupeptin). After incubation on ice for 15 min, cells were disrupted by sonication with two bursts, in 5 s pulses at 4°C. The suspension was fractionated by centrifugation at 1,000 ϫ g for 10 min at 4°C to remove nuclei and unbroken cells. The plasma membranes in the supernatant were separated by centrifugation at 83,000 ϫ g for 30 min at 4°C. The particulate extracts were resuspended in 0.1 ml of solubilization buffer (20 mM sodium phosphate buffer, pH 6.5, 10% glycerol, 1 mM CaCl 2 , and 0.5% CHAPS) at 4°C. The suspension was mixed by gentle pipetting, and the soluble proteins were fractionated by centrifugation at 175,000 ϫ g for 40 min at 4°C. The supernatant was removed and the pellet was extracted a second time with solubilization buffer by the same procedure. The first and second extracts were combined and constituted the detergent-soluble membrane fraction. About 50 g of protein was obtained from 25 ϫ 10 6 Y-79 cells.
Membranes from rat CG cells were prepared as above and with the following modifications. The homogenization buffer was 72 ml of 0.32 M sucrose in Tris-buffered saline containing protease inhibitors, for 360 ϫ 10 6 cells at days in vitro (DIV) 7 and 8. The plasma membranes in the homogenized cell suspension were fractionated by centrifugation at 18,000 ϫ g for 30 min at 4°C. Integral membrane proteins were fractionated with Triton X-114, as described previously (27). Briefly, the particulate membrane extract was resuspended in 18 ml of 1% Triton X-114 in Tris-buffered saline and incubated for 1 h at 4°C. The suspension was subjected to centrifugation at 10,000 ϫ g for 10 min at 4°C. The supernantant was warmed to 32°C for 10 min, and then the detergent and aqueous phases were separated by centrifugation. The detergent phase was mixed with an equal volume of Tris-buffered saline, incubated for 30 min at 4°C, and separated as described above. The detergent phase was diluted in cold 10 mM Tris acetic acid, pH 8, and 1 mM EDTA, and the detergent was removed by chromatography through a 2-ml Calbiosorb adsorbent column. Elution was performed with the dilution buffer, and fractions of 2 ml were collected in tubes containing 0.2 ml of 5% CHAPS. Fractions containing protein were pooled and constituted the detergent-soluble membrane fraction. About 700 -900 g of solubilized membrane protein were obtained from 360 ϫ 10 6 CG cells.
Preparation of PEDF-Ultralink TM Medium-Recombinant human PEDF at 2 mg/ml was dialyzed against coupling buffer containing 0.6 M sodium citrate, 0.1 M sodium carbonate, pH 9.0. A total of 5.5 ml of dialysate or of coupling buffer was transferred to a tube containing 0.25 g of dry Ultralink TM biosupport medium, and the manufacturer's instructions were followed. The coupling reaction was performed by incubation with gentle rotation at 25°C for 3 h. The suspension was then centrifuged at 1200 ϫ g for 10 min at 25°C, and the amount of uncoupled protein was determined from the supernatant. The amount of uncoupled protein ranged between 0.4 and 6.5% of the total protein added to the mixture. To block the remaining active groups in the coupled beads, 10 column volumes of 3 M of ethanolamine, pH 9, were added, and the mixture was rotated for 3 h at 25°C. The quenching solution was removed after centrifugation and the beads were washed with 15 ml of PBS, followed with 5 ml of 1 M NaCl. Finally, the beads were washed with 30 ml of PBS and stored in 0.05% sodium azide/PBS at 4°C. The PEDF affinity resin contained about 1-6 mg of PEDF per ml of resin.
PEDF Affinity Column Chromatography-PEDF affinity and control resins were packed in Polyprep chromatography columns (Bio-Rad) to yield a 0.25-ml settled bed volume and equilibrated with binding buffer (20 mM sodium phosphate buffer, pH 6.5, 150 mM NaCl, 10% glycerol, 1 mM CaCl 2 , and 0.5% CHAPS). Detergent-soluble membrane proteins from 25 ϫ 10 6 Y-79 cells were applied to the control column to absorb Ultralink TM -binding proteins. The flow-through and the first 7-column volume wash with the binding buffer were pooled and mixed with PEDF affinity beads (1.8 mg of PEDF/ml of beads) at a volume ratio of 7:1, and the mixture was rocked at 4°C for 16 h. The resin was transferred into a column and washed with 12 column volumes of binding buffer, and the bound proteins were eluted with 14 column volumes of low pH buffer (10 mM glycine, pH 2.0, 150 mM NaCl, 10% glycerol, 1 mM CaCl 2 , and 0.25% CHAPS) followed by 14 column volumes of 1 M NaCl in low pH buffer, 14 column volumes in high pH buffer (100 mM glycine, pH 11.0, 150 mM NaCl, 10% glycerol, 1 mM CaC1 2 , and 0.25% CHAPS), and finally with 1 M NaCl in high pH buffer. Proteins from each elution were precipitated with trichloroacetic acid.
Detergent-soluble membrane extract from 360 ϫ 10 6 CG cells was applied to a control resin column (0.5 ml) preequilibrated with 100 mM Tris-acetic acid, pH 8, 1 mM EDTA, and 0.5% CHAPS. The flow-through was mixed with 1 ml of PEDF beads (6 mg of PEDF/ml of Ultralink TM ) and incubated with gentle rotation at 4°C for 16 h. The PEDF affinity resin was transferred into a column and washed with 20 column volumes of 150 mM NaCl in binding buffer, and the proteins were eluted with 6 column volumes of low pH buffer (100 mM glycine, pH 2.5, 150 mM NaCl, 10% glycerol, 1 mM EDTA, and 0.5% CHAPS) followed by 6 column volumes of the same buffer at pH 11. Eluted proteins were concentrated by ultrafiltration with Centricon-30 (Amicon), following the manufacturer's protocol.

I-PEDF Induces Morphological Differentiation on Y-79
Cells-Before we could investigate the binding characteristics of a ligand to target cells, it was necessary to analyze its structural and biological properties. PEDF was chemically modified by iodination with 125 I and resolved by SDS-PAGE. Fig. 1 shows that 125 I-PEDF migrated as a single band corresponding to a 50-kDa protein, identical to that of unlabeled PEDF. Immunochemical analysis showed that the specific antibody to PEDF, Ab-rPEDF (12), recognized the radiolabeled protein on Western transfers or by immunoprecipitation, indicating that the chemical modification had not altered its main antigenic sites (data not shown). The radiolabeled protein was assayed for neurite outgrowth in human retinoblastoma Y-79 cell cultures (Fig. 1). We found that 125 I-PEDF, prepared from either native bovine or recombinant human PEDF protein, induced the typical morphological differentiation on retinoblastoma Y-79 cells as previously demonstrated for their unlabeled counterparts (12,13,24,26). Thus, 125 I-PEDF had the structural characteristics of the unlabeled protein and was obtained in a biologically active form.
Binding Analysis of 125 I-PEDF to Retinoblastoma Y-79 Cells-The binding of 125 I-PEDF to human retinoblastoma Y-79 cells was characterized in detail. The binding reaction mixtures were incubated at 4°C to avoid degradation of the ligand. PEDF (0.2-5 nM) was added to the cell cultures in defined media that had been conditioned for 16 h, to obtain conditions identical to those in the biological assay. A time course of 4.6 nM 125 I-PEDF binding to Y-79 cells indicated that specific binding was attained as early as 15 min after incubation and was maintained through the next 75 min (Fig. 2A).
The free and bound radioligand did not undergo major degradation and had an identical migration pattern by SDS-PAGE, even after incubation for 16 h, but the binding signal decreased in the presence of an excess of unlabeled PEDF, confirming the specificity of 125 I-PEDF binding to Y-79 cells (Fig. 2B). The specific binding augmented linearly with increasing concentrations of 125 I-PEDF in the range of 0 -1 nM, representing ϳ4% of the total radioactivity added to the cells (Fig. 2C). The amount of nonspecific binding was similar in the presence of an excess of unlabeled PEDF or without cells.
We then analyzed the PEDF binding profile to Y-79 cells to determine its physicochemical parameters. Binding was performed with a given amount of radioligand and increasing concentrations of unlabeled PEDF. Fig. 2D shows that Y-79 cells exhibited a saturable and specific binding of PEDF. Scatchard analysis of the binding data revealed a single class of binding sites (B max ϭ 271,200 sites per Y-79 cell) with an apparent dissociation constant (K d ) of 3.55 nM. A second and third experiment performed with different batches of Y-79 cells revealed similar kinetics, with K d ϭ 1.7-3 nM and B max ϭ 45,500 -97,000 sites per Y-79 cell (data not shown).
Effect of Antiserum to PEDF and Ovalbumin on the 125 I-PEDF Binding-We have previously shown that a rabbit polyclonal antiserum developed against recombinant human PEDF, Ab-rPEDF, reacts in a specific, sensitive, and linear fashion with bovine PEDF, and when preincubated with PEDF, it blocks its inducing-neurotrophic activity on human retinoblastoma Y-79 cells (12). To evaluate the effect of Ab-rPEDF on the l25 I-PEDF binding to receptors on Y-79 cells, the radioligand was preincubated with an excess of Ab-rPEDF before adding it to the cells binding reaction mixtures. We found that in the presence of Ab-rPEDF the 125 I-PEDF binding decreased to nonspecific binding levels (Fig. 3A), suggesting that Ab-rPEDF blocked the PEDF neurotrophic activity by preventing the binding of PEDF to receptors on Y-79 cells.
To determine whether serpins can compete for the 125 I-PEDF binding to Y-79 cells, we chose ovalbumin because it does not have inhibitory activity against proteases, which could interfere in the assay. In addition, it shares the lack of S 3 R conformational change upon cleavage of the serpin-exposed loop with PEDF (25, 28) but failed to induce neurite outgrowth in assays with Y-79 cells (data not shown). We found that in the presence of excess ovalbumin the binding of 125 I-PEDF to Y-79 cells was partially inhibited (Fig. 3B). Similar values were obtained when bound 125 I-PEDF was separated from free by centrifugation, and its radioactivity was determined after SDS-PAGE (data not shown).
Competition of 125 I-PEDF Binding with Synthetic Peptides 34-and 44-Mer-We have previously shown that recombinant PEDF polypeptide fragments with truncations from the carboxyl end retain neurotrophic activity (25). These results suggest that the smallest biologically active fragment, BA (posi-tions 44 -121 of human PEDF) contains the site that binds to cell surface receptor(s) on Y-79 cells. We have prepared two peptides from BA, peptides 34-mer (amino acid positions 44 -77) and 44-mer (amino acid positions 78 -121) to use as competitors of the 125 I-PEDF binding. As shown in Fig. 4, increasing concentrations of peptide 44-mer inhibited the 125 I-PEDF binding in a concentration-dependent fashion, with an EC 50 ϭ 5 nM. It is worth noting that the EC 50 of bovine PEDF was 3 nM, as calculated from data in Fig. 2D. Peptide 34-mer had an insignificant effect. Peptides 34-mer and 44-mer were used for neurite outgrowth assays with Y-79 cells, to evaluate their biological activity. After 22 days postattachment, only cells that had been treated with 44-mer showed elongation of processes similar to those for rPEDF-BH (24), the bacterially derived recombinant human PEDF (positions 44 -418) (Fig. 5). These results demonstrate that peptide 44-mer competed efficiently for the binding of PEDF to receptors on Y-79 and suggest that it contains the structural determinants for binding and neurotrophic activity of PEDF.
Cell Type Specificity of Cell Surface 125 I-PEDF Binding-We examined whether the binding was confined to Y-79 cells. Two human retinoblastoma cell lines (Y-79 and Weri), primary cultures of rat CG cells, mouse NIH3T3, hamster BHK, and monkey COS-7 cells were used in 125 I-PEDF binding reactions. The neuronal survival and differentiating effects of PEDF on the CG and retinoblastoma cells have been previously demonstrated (2-6), whereas the response of the other cells to PEDF stimuli is still unknown. Fig. 6 shows that the specific 125 I-PEDF binding for Y-79 and Weri cells reached 12 and 1.7%, respectively, whereas for the other cells, it was below 1%. The binding to NIH3T3 was the highest of the others, reaching 0.8% compared with CG and BHK cells with about 0.1% and no detection of binding to COS-7. These results suggest that the retinoblastoma cells have a higher number of PEDF-binding sites on their cell surface and/or binding sites with a higher affinity for PEDF than CG cells.
Binding Analysis of 125 I-PEDF to Cerebellar Granule Cell Neurons-CG cells, the most abundant neuronal subtype cells in the mammalian brain, undergo a maturation process during the course of culture, i.e., they can be distinguished between immature (DIV 0 -3) and mature (DIV 5ϩ) cells. We have observed that when added at specific DIVs, PEDF has a variety of rescuing effects on CG neurons, e.g. neuronal survival effect is observed after PEDF addition at DIV 1 (4), protection against glutamate neurotoxicity at DIV 8 (5) and antiapoptotic effect at DIV 2 but not at DIV 5 (6). We have determined the physicochemical characteristics of PEDF binding by CG cells cultured at different DIV. Fig. 7 shows that 125 I-PEDF bound to CG cells in a specific and saturable fashion, with binding parameters that were similar and independent of the age of the cultures. The ligand bound as a 50-kDa protein to about 1000 sites per CG cell, with a K d of 3 nM. The K d value is in agreement with the half-maximum dose for the PEDF biological activity on CG cells, 0.5-3 nM (4,5), and is similar to that of Y-79 cells (see Fig.  2D). However, the number of PEDF-binding sites were Ͼ100fold lower than those for Y-79 cells (see Fig. 2D). These results suggest that PEDF binds to a similar cell surface protein on immature and mature CG cells and on retinoblastoma cells.
PEDF Affinity Column Chromatography of Y-79 Retinoblastoma and CG Cells Membrane Proteins-To isolate the cell surface protein(s) with binding affinity for PEDF, plasma membrane proteins of Y-79 retinoblastoma and CG cells were extracted and solubilized with detergents, CHAPS, or Triton X-114. Triton X-114 was used to isolate integral membrane proteins (27). Preliminary experiments showed that the detergent-soluble membrane extracts contained the PEDF binding activity, e.g. 0.5-7.5 fmol of 125 I-PEDF specifically bound to detergent-soluble membrane proteins of 0.2-2 ϫ 10 6 Y-79 cells (radioligand at 2.4 nM). Membrane proteins with affinity for PEDF were fractionated by PEDF affinity column chromatography (Fig. 8). Most of the solubilized membrane proteins from Y-79 cells did not bind to PEDF; however, a main protein with an apparent molecular weight of 80,000 bound and eluted from the column with buffers of elevated pH (Fig. 8A). Another band of less intensity and migrating as a molecular weight 60,000 protein was also detected in the eluates. Biotinylation of Y-79 cell surface proteins followed by PEDF affinity chromatography showed that a PEDF-binding protein of 80 kDa is a cell surface protein (data not shown). Similarly, a CG cell mem- was added to BHK, COS-7, NIH3T3, CG, and human retinoblastoma Weri and Y-79 cell cultures in the absence (total binding) or presence of a 66-fold molar excess of unlabeled PEDF (nonspecific binding). The bound/total was calculated by dividing the specific binding by the total radioactivity added per assay and expressed per 500,000 cells. All experimental points are given as the average of duplicates for CG cells and triplicates for the others. brane protein with a molecular weight of about 80,000 bound to the column and eluted with buffers at pH 11 (Fig. 8B). The level of purity of the eluted fractions seemed to be a reflection of the number of PEDF-binding sites (B max ) for Y-79 and CG cells (see Figs. 2 and 7). The size and biochemical characteristics of the binding and elution of the PEDF-binding proteins from human retinoblastoma and rat CG neurons suggest that they are homologous proteins. Thus, we have identified a membrane protein from human retinoblastoma and rat CG cells with an apparent molecular weight of ϳ80,000 that has PEDF-binding properties and is consistent with a PEDF receptor. DISCUSSION Several lines of evidence implicate inhibition of proteases in serpin biological activities. However, PEDF has no known protease target for inhibition, and furthermore, its neurotrophic activity is determined by sequences away from its serpin-ex-posed loop, the structural determinant of serpin protease inhibition (25). The present study indicates that PEDF can interact with cells derived from the retina and the central nervous system. The data suggest that PEDF has a receptor on the surface of cells that respond to its stimuli. This conclusion is supported by several lines of evidence: 1) biologically active 125 I-PEDF binds directly to both human retinoblastoma and rat CG cells; 2) 125 I-PEDF binding is competed by native and recombinant PEDF and the biologically active peptide 44-mer; 3) 125 I-PEDF binding is inhibited by Ab-rPEDF previously shown to block the PEDF biological activity; 4) PEDF binding affinity to receptors on retinoblastoma cells is similar to that of CG cells; and 5) the biochemical characteristics of a PEDFbinding protein from membranes of retinoblastoma are similar to those of CG cells. It is clear that further information is required to confirm that this binding activity and PEDF-binding protein is a PEDF receptor that, upon its interaction with its ligand, triggers the necessary signal transduction events for neurotrophic activity. However, the present study indicates that the first step in the biological events of PEDF is established as binding to receptors on the surface of target cells.
The data reported here also provide pharmacological and biochemical evidence that PEDF and peptide 44-mer bind to the same site, called PEDF receptor, on Y-79 cells: the specific and saturable PEDF binding to cells is blocked by peptide 44-mer in a concentration-dependent fashion. In addition, this peptide is biologically active, e.g. exhibits neuronal differentiating activity on human retinoblastoma cells (Fig. 4) and is protective for motor neurons against chronic glutamate toxicity in organotypic spinal cord cultures (8). These observations suggest that the region spanning positions 78 -121 is the site in PEDF that binds to receptors on Y-79 cells. The region of 44-mer is conserved among the human, bovine, and mouse PEDF sequences. Sequence comparison and alignment using a BLAST program (29) of the 44-mer and GenBank TM data bases revealed that the highest similarity is found for regions toward the amino-terminal end of other serpins (35 Ϯ 10% identity), including ovalbumin and protease nexin-1. Sequence similarity could account for the partial inhibition of PEDF binding by ovalbumin. In a previous study, we analyzed the heparin-binding site of PEDF and proposed a spatial structural model for PEDF (16). In this model, the 44-mer region, the receptor- FIG. 7. PEDF binding to cell surface receptors on CG cells. Radioligand was prepared from recombinant human PEDF. A, cultures of CG cells at DIV 9 were incubated with 2.9 nM 125 I-PEDF at 4°C for 16 h. Medium was removed, and the cells were washed with binding buffer and lysed with 1% SDS. The proteins were precipitated with tricholoroacetic acid before resolving by SDS-PAGE, followed by autoradiography. Samples were applied to a 10 -20% polyacrylamide gel as follows: lane 1, medium; lane 2, CG cells. Binding assays were performed two times, and a photograph of a representative x-ray film is shown. Migration of molecular weight standards is indicated to the right. B and D, photographs of CG cells at DIV 2 and 7, respectively, visualized under light microscopy. C and E, saturation and Scatchard analysis of PEDF binding by CG cells at DIV 2 and 7, respectively. Binding was performed with 0.8 nM 125 I-PEDF and increasing concentrations of unlabeled human PEDF. Data was analyzed by linear regression with Microsoft Excel 97 to obtain Scatchard binding plots. The following values were obtained: for CG cells at DIV 2 (C), K d ϭ 4.1 nM and B max ϭ 1187 sites/cell; for CG cells at DIV 7 (E), K d ϭ 2.25 nM and B max ϭ 1000 sites/cell. Similar results were obtained by nonlinear regression, and Scatchard analyses were obtained using GraphPad Prism software. Experimental points are given as the average of duplicates for DIV 2 and triplicates for DIV 7. Analyses were performed more than three times, and a photograph of a representative silverstained gel for each preparation is shown. Migration of molecular weight standards is indicated to the left. The arrow and asterisk point to the migration of proteins with a molecular weight of 80,000 and 60,000, respectively. binding site, is located in a distinct area and away from the heparin-binding site. It will be necessary, however, to isolate cDNA clones for PEDF receptor and express PEDF receptor in a heterologous cell to provide evidence that peptide 44-mer and PEDF bind to it.
Previous reports have demonstrated two different receptors for serpins, the low density lipoprotein receptor-related protein and the serpin-enzyme complex receptor. The low density lipoprotein receptor-related protein, a multifunctional cell surface receptor that binds and endocytoses several distinct ligands including serpin-enzyme complexes, is the hepatic receptor responsible for the clearance of serpin-enzyme complexes (30 -32). Interestingly, its gene is abundantly expressed in most neurons of the central nervous system, particularly in the cerebellum (33). In addition, the low density lipoprotein receptor-related protein can mediate the neurite outgrowth activity of apolipoprotein E in a central nervous system-derived neuronal cell line (34,35). These observations suggest that the low density lipoprotein receptor-related protein contains the necessary biochemical components for signaling the transduction of neurite outgrowth upon binding to a ligand and becomes a candidate receptor for PEDF. However, the low density lipoprotein receptor-related protein is unable to bind to immobilized native or proteolytically modified forms of serpins antithrombin III, heparin cofactor II and ␣1-antitrypsin, suggesting that it would be unable to bind PEDF (30). In addition, retinoblastoma cells do not have the ligand-binding subunit of the low density lipoprotein receptor-related protein (585-kDa). 2 These observations exclude the possibility that the PEDF binding activity described here is directed toward the low density lipoprotein receptor-related protein.
On the other hand, the serpin-enzyme complex receptor has a ligand-binding subunit of about 78 kDa, binds several serpinenzyme complexes and soluble amyloid-␤ peptide, and is also present in neuronal cells (36 -38). Several observations suggest that the serpin-enzyme complex receptor might not be related to the PEDF receptor: 1) proteolytically modified ␣1-antytrypsin competes for binding of serpin-enzyme complexes to the serpin-enzyme complex receptor and vice versa, whereas the purified 51-kDa amino-terminal fragment of ␣1-antitrypsin does not compete for binding of serpin-enzyme complex (39); and 2) on the other hand, an amino-terminal fragment of PEDF, peptide 44-mer, competes for the binding of 125 I-PEDF to Y-79 cells (Fig. 4A). It has been shown that the serpinenzyme complex receptor recognizes a pentapeptide domain in the carboxyl-terminal fragment of ␣1-antitrypsin-protease complex (FVFLM, amino acid positions 370 -374) and in the amyloid-␤ peptide (IIGLM, positions 31-35) (38,40). We have compared these sequences to the 44-mer and have not been able to identify significant similarity. Thus, the present studies cannot reveal the identity of the PEDF receptor, but they provide evidence that PEDF acts to promote neuronal differentiation and survival on retinoblastoma and CG cells via a novel receptor-mediated mechanism, which may indicate a novel role for serpins in neurotrophic activity.
The results of this study may also have important implications for neurodegenerative diseases. In the vertebrate retina, the PEDF gene is expressed by retinal pigment epithelial cells (14,18), and the protein product is secreted from these cells and localized in the interphotoreceptor matrix (12,15). The PEDF gene maps next to loci for several inherited retinal dystrophies, such as retinitis pigmentosa, progressive cone dystrophy, and central areolar choroidal dystrophy (41)(42)(43)(44). We have recently shown that intravitreal injections of PEDF in animal models for retinitis pigmentosa resulted in a delay of photoreceptor cell degeneration, implying a survival effect on photoreceptor cells in vivo (10). It is not known yet whether PEDF acts directly on receptors on the photoreceptor cells or via receptors on other retinal cells, which will eventually have an effect on the photoreceptor cells. The reported neurotrophic activities on CG cells in vitro also point out important implications for cerebellar neurodegenerative diseases, e.g. xeroderma pigmentosum and Cockayne syndrome, hereditary disorders characterized by impaired DNA repair and neurodegeneration, in particular apoptosis of CG neurons (45). Although the protein is detected in cerebrospinal fluid (8), we have not detected it in the CG cell culture media, 3 suggesting that PEDF could be a paracrine neurotrophic factor for cerebellar neurons. Taken together with these data, the present study reveals the binding of PEDF to its cell surface receptor and constitutes the first step in understanding its mechanism of action as a potential neurotrophic factor for the retina and central nervous system.