Identification of a Novel Inhibitor of the Canonical Wnt Pathway

Abstract

Wnt signaling is known to regulate multiple processes including angiogenesis, inflammation, and fibrosis. Here, we identified a novel inhibitor of the Wnt pathway, pigment epithelium-derived factor (PEDF), a multifunctional serine proteinase inhibitor. Both overexpression of PEDF in transgenic mice and administration of PEDF protein attenuated Wnt signaling induced by retinal ischemia. Furthermore, PEDF knockdown by small interfering RNA (siRNA) and PEDF knockout in PEDF^−/− mice induced activation of Wnt signaling. PEDF bound to LRP6, a Wnt coreceptor, with high affinity (K_d [dissociation constant] of 3.7 nM) and blocked the Wnt signaling induced by Wnt ligand. The physical interaction of PEDF with LRP6 was confirmed by a coprecipitation assay, which showed that PEDF bound to LRP6 at the E1E2 domain. In addition, binding of PEDF to LRP6 blocked Wnt ligand-induced LRP6-Frizzled receptor dimerization, an essential step in Wnt signaling. These results suggest that PEDF is an endogenous antagonist of LRP6, and blocking Wnt signaling may represent a novel mechanism for its protective effects against diabetic retinopathy.

INTRODUCTION

Diabetic retinopathy (DR) is a common complication of diabetes, predominantly affecting the capillaries in the retina (1, 3). The progressive pathological changes of DR include loss of pericytes, basement membrane thickening, leukostasis, increased capillary permeability, and retinal neovascularization (NV) (3, 45, 46, 62). It has been shown that oxidative stress and chronic inflammation play important roles in DR (39). Our recent studies showed that the Wnt pathway is activated in the retina of diabetic patients and diabetic animal models (13, 18). Furthermore, blocking the Wnt pathway with a specific inhibitor of attenuated retinal inflammation, vascular leakage, and retinal NV (18). These findings suggest that the Wnt pathway plays a key pathogenic role in the retinal inflammation and NV observed in DR.

The Wnt signaling pathway is a conserved intracellular signaling pathway (31, 40, 65). Binding of Wnt ligands such as Wnt3a to a coreceptor complex consisting of the Frizzled (Fz) receptor and low-density lipoprotein receptor-related protein 5 (LRP5) or LRP6 initiates an intracellular signaling cascade (13). This cascade includes the association of the Wnt receptor with the Dishevelled protein (Dsh), promoting the interaction between Dsh and Axin (73). The Dsh-Axin association triggers the disruption of the destructive complex, which is composed of adenomatous polyposis coli (APC) Axin, casein kinase-1α (CK1α), and glycogen synthase kinase 3 (GSK3) (20). Consequently, GSK3 does not phosphorylate β-catenin (24), releasing it from the Axin complex and leading to β-catenin accumulation and nuclear translocation. The nuclear β-catenin subsequently dimerizes with T-cell factor (TCF) and activates the transcription of multiple target genes, including the genes encoding some inflammatory, fibrogenic, and angiogenic factors (55, 71).

The Wnt/β-catenin pathway is known to regulate multiple cellular processes such as cell differentiation, inflammation, carcinogenesis, fibrosis, and angiogenesis (11, 29, 42). It has been shown that TCF/β-catenin regulates vascular endothelial growth factor (VEGF) gene transcription (61, 81). Seven TCF/β-catenin-binding sites have been identified in the promoter region of the VEGF gene (21, 55). Furthermore, knockout of very-low-density lipoprotein receptor (VLDLR), a negative regulator of the Wnt pathway, results in activation of the Wnt pathway and overexpression of VEGF, retinal inflammation, and subretinal NV (16, 17, 32).

Pigment epithelium-derived factor (PEDF) is a 50-kDa glycoprotein, belonging to the serine proteinase inhibitor (SERPIN) superfamily (7, 67). PEDF has been found to be a potent inhibitor of angiogenesis (12, 23). It has also been reported that PEDF has broad beneficial effects including anticancer, antioxidation, anti-inflammation, and antifibrosis activities (23, 44, 70, 80). Furthermore, PEDF levels are decreased in the aqueous and vitreous fluids from patients with DR and in the retinae of DR animal models (22, 52, 53). Therefore, decreased PEDF levels have been suggested to contribute to DR (9, 10, 27, 64). Despite extensive studies, the mechanism(s) underlying the broad biological activities of PEDF has not been fully elucidated. Although previous studies showed that PEDF binds to several potential cell surface receptors such as glycosaminoglycans, collagens on the surface of retinoblastoma cells, and phospholipase A₂ξ (PLA₂ξ) (49), binding to these receptors cannot explain the broad effects of PEDF.

Here, we have studied the interactions of PEDF with the Wnt pathway in vivo and in vitro and demonstrated that PEDF is an endogenous inhibitor of the canonical Wnt pathway.

MATERIALS AND METHODS

Plasmid construction and protein expression.

The expression vectors of the extracellular region of human LRP6 (lacking the C terminus and transmembrane domain) tagged with Myc (LRP6N-Myc), the extracellular region of LDLR tagged with Myc (LDLRN-Myc), and the cysteine-rich domain (CRD) of the Fz8 receptor tagged with an immunoglobulin gamma Fc epitope (Fz8-CRD-IgG) were constructed as described previously (57). Plasmids expressing LRP6 and VLDLR were constructed by cloning the full-length human LRP6 and VLDLR cDNAs into the pcDNA3 vector (Invitrogen, Carlsbad, CA). The E1E2 and E3E4 domains of LRP6 were fused with a His tag sequence, subcloned into the pET28b(+) vector (Novagen, Gibbstown, NJ), and expressed in Escherichia coli BL21(DE3) by isopropyl-β-d-thiogalactopyranoside (IPTG) induction. For purification of the LRP6 E1E2 or E3E4 domains, the inclusion body was isolated and denatured in a buffer containing 6 M guanidine hydrochloride. The peptides were purified by passing the solubilized inclusion body proteins through a Ni-resin column (Novagen, Gibbstown, NJ). In-column refolding of the peptides was achieved by serial changes of buffers: 10 column volumes (CV) of buffer A (0.1% Triton X-100, glutathione/glutathione disulfide [GSH/GSSG] in phosphate-buffered saline [PBS], pH 7.8), 10 CV of buffer B (5 mM beta-cyclodextran, GSH/GSSG in PBS, pH 7.8), and washing buffer (20 mM imidazole in PBS, pH 7.8). The peptides were eluted from the column with elution buffer (Novagen, Gibbstown, NJ). The extracellular domain of human VLDLR (VLDLRN) was expressed and purified following the same procedure as above. The eluted fractions were dialyzed in PBS for 24 h, and the peptide concentration was measured by Bradford assay. The purity of the recombinant peptides was determined by SDS-PAGE and Coomassie blue staining.

Cell culture, transfection, and transduction.

ARPE19, CHO, and HEK293T cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum, antibiotics, and glutamine. Human retinal endothelial cells were purchased from Cell Science (Canton, MA) and cultivated as described by the supplier and grown on culture dishes precoated with 0.2% gelatin (Sigma, St. Louis, MO). The cells were used in low passages. Human Müller cells (MIO-MI) were cultured in DMEM (Mediatech Inc., Manassas, VA) supplemented with 10% fetal bovine serum (FBS). Human retinal pericytes were a kind gifted from Timothy Lyons and were grown in EBM-2 endothelial basal serum-free culture medium and EGM-2-MV (microvascular endothelial cell growth medium) SingleQuots growth supplement (Clonetics, Walkersville, MD), including 5% fetal bovine serum, 0.1% human epithelial growth factor (hEGF), 0.04% hydrocortisone, 0.1% VEGF, 0.4% human fibroblast growth factor B (hFGF-B), 0.1% revitropin insulin-like growth factor (R³-IGF-1), 0.1% ascorbic acid, and 0.1% GA-1000. For stimulation of Wnt signaling, ARPE19 or HEK293T cells were cultured overnight in six-well plates, serum starved in DMEM for 6 h, and incubated with the conditioned medium (CM) from cells stably expressing Wnt3a (Wnt3a-conditioned medium [WCM] in the presence of different concentrations of PEDF. For transient transfection, ARPE19 cells were transfected with the plasmids using Lipofectamine 2000 (Invitrogen, Carlsbad, CA), and CHO and HEK293T cells were transfected using Fugene 6 (Roche, Indianapolis, IN) according to the manufacturer's protocol. For infection with recombinant adenovirus expressing a constitutively active mutant form of β-catenin (Ad-S37A), ARPE19 cells in six-well plates were seeded in growth medium and cultured for 24 h. The cells were infected at the desired multiplicity of infection (MOI) of Ad-S37A or Ad-β-Gal virus as previously described (72).

PEDF protein purification and activity.

Serum-free medium from the cell line stably expressing human PEDF was gently harvested. The collected medium was precipitated with 80% ammonium sulfate suspended in buffer A (20 mM Na-phosphate, 150 mM NaCl). The suspended material was dialyzed against 50 mM phosphate buffer B (50 mM NaCl, 1 mM dithiothreitol, and 10% glycerol, pH 6.2). The dialyzed PEDF was further purified by cation-exchange chromatography. After the column was washed with the equilibrating buffer B, PEDF was eluted with a linear gradient from 50 mM to 500 mM NaCl. Samples from the collected fractions were subjected to 8% SDS-PAGE. One gel was stained with Coomassie blue, and the other one was applied to Western blot analysis with a rabbit anti-PEDF antibody. To determine PEDF activity, bovine retinal capillary endothelial cells (BRCEC) were seeded in a gelatin-coated 96-well plate. The plates were subjected to hypoxia in the presence of various concentrations of PEDF. The viable cells were quantified using an MTT [3-(4,5-dimethylthiazol-2-yl)2 2,5-diphenyl tetrazolium bromide] assay kit (Roche, Indianapolis, IN).

Luciferase reporter activity assays.

HEK293T cells were plated at 2 × 10⁵ cells/well in 24-well plates 24 h prior to transfection. Cells were transfected with TopFlash (100 ng), pRL-TK (5 ng), and LRP6 (20 ng) expression vectors using Fugene 6 (Roche, Indianapolis, IN) according to the manufacturer's protocol. As needed, a negative- or positive-control vector was added at identical DNA amounts. At 24 h posttransfection, the cells were incubated with 500 μl of L-conditioned medium (LCM) or WCM and different concentrations of PEDF. The cells were cultured for another 16 h and lysed in 100 μl of passive lysis buffer (Promega, Madison, WI). The firefly luciferase (TopFlash) activity and Renilla (pRL-TK) luciferase activity were measured using a dual luciferase assay kit (Promega, Madison, WI) in a luminometer (Berthold, Oak Ridge, TN) following the manufacturers' protocols. Renilla luciferase activity was measured as an internal control for transfection efficiency. The Wnt-induced TopFlash activity is depicted as the ratio of the activity of firefly to Renilla luciferase. All assays were performed in triplicate and repeated in three independent experiments.

Real-time reverse transcription-PCR (RT-PCR).

Total RNA was extracted using TRIzol reagent (Invitrogen, Carlsbad, CA). The cDNA was synthesized from 1 μg of total RNA using TaqMan Reverse Transcription Reagents according to the manufacturer's protocol (Applied Biosystems, Foster City, CA), and the reaction product was subjected to PCR amplification using a MyiQ Bio-Rad thermal cycler (Bio-Rad, Hercules, CA). The following primers were used for the PCR: for cyclin D1, 5′-CCGTCCATGCGGAAGATC-3′ (forward) and 5′-ATGGCCAGCGGGAAGAC-3′ (reverse); for c-Myc, 5′-GCCACGTCTCCACACATCAG-3′ (forward) and 5′-TCTTGGCAGCAGGATAGTCCTT-3′ (reverse); for connective tissue growth factor (CTGF), 5′-TGCGTTTTGGAGCTAGCGGACCA-3′ (forward) and 5′-CGAGGACCATACAGCACGTGCCAG-3′ (reverse); and for glyceraldehyde-3-phosphate dehydrogenase (GAPDH), 5′-ATGGTGAAGGTCGGTGTGAAC-3′ (forward) and 5′-GTGCCGTTGAATTTGCCGTGA-3′ (reverse).

Confocal microscopy.

Lysine residues in PEDF and bovine serum albumin (BSA) were labeled by fluorescein isothiocyanate (FITC) using an FITC labeling kit (Invitrogen, Carlsbad, CA). CHO cells grown on coverslips were transfected with an LRP6-Myc or VLDLR-Myc expression plasmid and incubated with FITC-PEDF or FITC-BSA for 2 h. The cells were then washed in cold PBS, fixed in 4% paraformaldehyde in PBS for 30 min, permeabilized with 0.1% Triton X-100 in PBS for 10 min, and blocked with 5% goat serum and 3% albumin in PBS for 15 min. After the cells were washed, they were incubated for 2 h with a mouse antibody for LRP6 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA). The cells were washed and incubated with a Texas Red-conjugated goat anti-mouse secondary antibody (Molecular Probes, Eugene, OR) in PBS with 5% goat serum and 3% albumin for 2 h, washed, and mounted on slides with 4′,6′-diamidino-2-phenylindole (DAPI; Vector, Burlingame, CA). Confocal images were collected using a Leica SP2 MP confocal microscope and analyzed using the Leica confocal software (Leica, Bannockburn, IL).

OIR mice, Vldlr^−/− mice, PEDF-TG mice, and PEDF^−/− mice.

All of the animal experiments were performed in compliance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. C57BL6 mice and Vldlr^−/− mice were purchased from Jackson Laboratories (Bar Harbor, ME). PEDF^−/− mice were provided as a generous gift by S. J. Wiegand from Regeneron Pharmaceuticals, Inc. (Tarrytown, NY) (34). Mice overexpressing PEDF in the retina (PEDF transgenic [PEDF-TG]) were generated using the human PEDF cDNA under the control of the chicken β-actin promoter. The oxygen-induced retinopathy (OIR) model was induced following an established protocol (63). Purified PEDF protein (82) was injected into the vitreous of the right eye (10 μg/eye) of wild-type (WT) mice with OIR at postnatal day 14 (P14) and into the Vldlr^−/− mice (Jackson Laboratories, Bar Harbor, ME) at P60. The same dose of BSA was intravitreally injected into the contralateral eyes as a control. The retinas or eye cups were collected 2 days postinjection for Western blot analysis.

Enzyme-linked immunosorbent assay (ELISA).

ARPE19 cells were plated in six-well plates and grown until 70 to 80% confluence. For induction of the Wnt signaling pathway in vitro, ARPE19 cells were treated with WCM only, WCM plus PEDF, or LCM as a negative control. For LiCl treatment, the cells were treated with 20 mM LiCl (Sigma, St. Louis, MO) for 24 h. The same volume of the conditioned medium from the treated cells or infected cells was used for a human VEGF-specific enzyme-linked immunosorbent assay (ELISA) using a VEGF ELISA kit (R&D Systems Inc., Minneapolis, MN) and following the manufacturer's protocol.

Cell surface receptor-ligand binding assay and affinity measurement.

For binding assays, PEDF and BSA were labeled with FITC using an FITC labeling kit (Invitrogen, Carlsbad, CA) following a protocol recommended by the manufacturer. CHO cells stably expressing LRP6-Myc or VLDLR-Myc were grown in 96-well plates (in triplicate) and incubated with different concentrations of FITC-PEDF or FITC-BSA for 2 h at 4°C. After the cells were washed with cold PBS three times, the fluorescence on the cells was measured in a Wallac fluorometer (Perkin Elmer, Waltham, MA). Saturable binding curves were plotted using Microsoft Excel software by fitting the bound FITC-PEDF intensities. Scatchard analysis was performed by plotting the ratios of the bound FITC-PEDF to the free FITC-PEDF against the concentrations of the bound FITC-PEDF in each experiment. All data were analyzed in Microsoft Excel.

Western blotting, cellular fractionation, and immunoprecipitation.

Cells were washed with cold PBS and lysed in radioimmunoprecipitation assay (RIPA) buffer (150 mM NaCl, 1% NP-40, 0.1% SDS, 50 mM Tris-HCl, pH 8.0, 1 mM EDTA, and 1 mM phenylmethylsulfonyl fluoride [PMSF]). For in vivo assays, the free-floating eye cups were homogenized in 150 μl of ice-cold tissue lysis buffer (50 mM Tris-HCl, pH 7.8, 5 mM EDTA, 0.1% SDS, 1% NP-40, 2.5% glycerol, 100 mM NaCl, and 1 mM fresh PMSF). The lysates were cleared by centrifugation at 12,000 × g for 20 min at 4°C, and the protein concentration of the lysates was determined using a Bradford assay. Proteins (50 μg) were resolved by 8 to 12% SDS-PAGE and transferred onto nitrocellulose membranes. Rabbit polyclonal antibodies against β-catenin, Myc tag, G_α12, and VEGF as well as a monoclonal anti-LRP6 antibody were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). The polyclonal antibody for phosphorylated LRP6 (pLRP6) was purchased from Cell Signaling (Danvers, MA). The signal intensity was quantified using SynGene tool imager software (SynGene, Frederick, MD).

For cellular fractionation, the nuclear fraction was isolated using a Fractionation System Kit (Biovision, Mountain View, CA), and the nuclear protein concentration was measured using a Bradford assay. Equal amounts of nuclear proteins were blotted with a polyclonal antibody for β-catenin and then reblotted with a mouse antibody for TATA-binding protein (TBP) (Abcam, Cambridge, MA) or lamin B (Santa Cruz Biotechnology, Inc., Santa Cruz, CA).

For immunoprecipitation (IP), the cells were lysed in phosphate-IP buffer (150 mM NaCl, 2 mM EDTA, 1% NP-40, 50 mM Tri-HCl, pH 7.4, 0.1% SDS, and 1 mM PMSF). Cellular lysates or conditioned medium was incubated with the appropriate antibodies at 4°C overnight, followed by the addition of protein A-Sepharose beads (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) for 2 h at 4°C. The beads were then washed with washing buffer containing 150 mM NaCl, 1 mM EDTA, 1% NP-40, 10 mM Tri-HCl, pH 7.4, and 1 mM PMSF, and the precipitated proteins were subjected to SDS-PAGE, followed by immunoblotting with the appropriate antibodies.

siRNA knockdown of endogenous PEDF.

A small interfering RNA (siRNA) transient transfection was performed using Lipofectamine 2000 (Invitrogen, Carlsbad, CA) according to the manufacturer's instruction. The human PEDF siRNA sequence used was 5′-GGAAAUUCCCGAUGAGAUCTT-3′ (Ambion, Inc., Austin, TX). The cells were treated with either LCM or WCM and lysed in 100 μl of passive lysis buffer (Promega, Madison, WI) 16 h after administration of LCM or WCM. For the TCF/β-catenin transcriptional activity assay, 24 h after the siRNA transfection, the cells were cotransfected with the TopFlash plasmid (firefly luciferase) and pRL-TK (Renilla luciferase) for 24 h. For the VEGF ELISA, 24 h after the transfection with siRNA targeting PEDF (siPEDF), the cells were treated with LCM or WCM, and then the secreted VEGF was measured in the culture medium.

Biotinylation assay using sulfo-NHS-LC-biotin.

ARPE19 cells were treated with WCM alone or WCM plus PEDF. The cells were incubated with 0.5 mg/ml sulfo-NHS-LC-biotin (sulfosuccinimidyl-6-biotinamido-hexanoate; Pierce, Rockford, IL) for 30 min. To exclude nonspecific binding of biotin from cell surface proteins, cells were treated with 100 mM glycine solution in ice-cold phosphate-buffered saline at pH 8.0. After biotinylated proteins were quenched, the cells were lysed in 0.2 ml of RIPA lysis buffer. The lysates were immunoprecipitated with the anti-LRP6 antibody, and the immunoprecipitates were analyzed by immunoblotting with the antibiotin antibody (Sigma, St. Louis, MO), anti-LRP6 antibody (Santa Cruz Biotechnology, Inc., Santa Cruz, CA), and anti-cadherin-P antibody (Santa Cruz Biotechnology, Inc., Santa Cruz, CA). Alternatively, the lysates were precipitated with immobilized NeutrAvidin protein (Pierce, Rockford, IL) and blotted with anti-LRP6 antibody, antibiotin antibody, or anti-cadherin-P antibody.

RESULTS

PEDF attenuates β-catenin nuclear translocation and VEGF overexpression in oxygen-induced retinopathy (OIR) and Vldlr^−/− mouse retinae.

To determine whether PEDF can modulate the Wnt signaling pathway, we used two different ocular NV models: OIR mice, which develop ischemia-induced retinal NV (63), and Vldlr^−/− mice, which develop subretinal NV (16, 32, 36). Our previous studies showed that the Wnt pathway is activated in the retinae of both of the models (16). Recombinant human PEDF purified from a stable cell line expressing PEDF (73) was injected intravitreally into the eyes of OIR mice at postnatal day 14 (P14). As shown by Western blot analysis, the nuclear β-catenin levels were significantly higher in the OIR retina at P16 than in the age-matched normal control (18). PEDF injection blocked the increase of the nuclear β-catenin levels in the OIR retina, compared to the BSA-injected contralateral control retinas (P < 0.01) (Fig. 1A).

Fig. 1.

Inhibition of β-catenin nuclear translocation and VEGF overexpression by PEDF in the retinae of OIR and Vldlr^−/− mice. (A) Purified PEDF was injected intravitreally into the eyes (0.5 and 1.0 μg/eye) of OIR mice at age P14, with BSA (1.0 μg/eye) as a control. At P16, retinal levels of nuclear β-catenin and total VEGF were measured by Western blot analysis. Total VEGF and the nuclear β-catenin levels were quantified by densitometry and normalized by β-actin and lamin B levels, respectively, and expressed as a percentage of that in normal and WT controls (mean ± standard deviation; n = 5). *, P < 0.05; **, P < 0.01 (compared to the BSA control). (B) PEDF was injected intravitreally into the right eyes (10 μg/eye) of 2-month-old Vldlr^−/− mice and the same amount of BSA into the left eyes. Two days after the injection, nuclear β-catenin levels and total VEGF were measured in the eye cups using Western blot analysis. Nuclear β-catenin levels were quantified by densitometry and normalized by lamin B levels and expressed as a percentage of the WT control value (mean ± standard deviation; n = 6). **, P < 0.01.

PEDF was also injected into the vitreous of Vldlr^−/− mice at P60. Western blot analysis at 48 h postinjection showed that PEDF decreased the nuclear β-catenin levels in Vldlr^−/− eye cups by more than 50% compared to levels of the BSA-injected contralateral controls (P < 0.01) (Fig. 1B). Furthermore, the expression of VEGF, a target gene regulated by TCF/β-catenin, was also significantly downregulated by PEDF injection in the retinae of the OIR and Vldlr^−/− models, consistent with the changes in the nuclear β-catenin levels (Fig. 1). These results indicate that the administration of PEDF suppresses the activation of the Wnt pathway in the retinae of the OIR and Vldlr^−/− models.

The Wnt pathway is inhibited by PEDF overexpression in PEDF-TG mice and activated in PEDF^−/− mice.

To verify the modulation of Wnt signaling by endogenous PEDF, we generated PEDF-TG mice which overexpress PEDF in the retina. PEDF-TG and wild-type (WT) mice were exposed to 75% oxygen from P7 to P12 to generate OIR. As shown by Western blot analysis, OIR decreased PEDF levels in the WT retina, but not in the PEDF-TG mice, at P16 (Fig. 2A). Furthermore, OIR increased nuclear β-catenin levels and induced VEGF expression by 1.8- and 1.6-fold, respectively, in the WT retina but induced smaller increases in the PEDF-TG retina (Fig. 2A). This result suggests that high levels of PEDF attenuate the ischemia-induced accumulation of β-catenin and the overexpression of VEGF.

Fig. 2.

Wnt signaling in the eye cups of PEDF-TG and PEDF^−/− mice with OIR. (A) WT and PEDF-TG mice were exposed to 75% oxygen from P7 to P12 and then returned to room air. As a control, another group of age-matched WT and PEDF-TG mice were maintained in constant room air. At P16, the mice were euthanized, and then the retinae were dissected and homogenized. Nuclear levels of β-catenin and lamin B and total levels of VEGF, PEDF, and β-actin were measured by Western blot analysis using equal amounts of retinal proteins from each mouse. Each lane represents an individual mouse. (B) Using the same conditions as for the OIR-PEDF-TG mouse model, PEDF^−/− mice were subjected to OIR. The nuclear levels of β-catenin and lamin B and total levels of VEGF, PEDF, and β-actin in the retinae of WT and PEDF^−/− mice with OIR or under normoxic conditions were determined using Western blot analysis. (A and B) Retinal levels of the nuclear β-catenin and total VEGF in panels A and B were quantified by densitometry and normalized by lamin B and β-actin levels, respectively, and expressed as a percentage of the WT control (mean ± standard deviation; n = 6). *, P < 0.05; **, P < 0.01.

PEDF^−/− mice were used to confirm the role of endogenous PEDF in the regulation of Wnt signaling. The PEDF^−/− retinae with OIR showed significantly higher induction of the nuclear β-catenin and VEGF levels than PEDF^−/− mice under normal conditions. These increases were significantly higher than the increase in the WT retinae of OIR mice. This suggests that the lack of PEDF enhances the nuclear β-catenin accumulation and VEGF overexpression in OIR (Fig. 2B).

Inhibition of the Wnt ligand-induced Wnt pathway activation by PEDF.

To further confirm a direct modulation of Wnt signaling by PEDF, we evaluated the effect of PEDF on Wnt ligand (Wnt3a)-mediated Wnt signaling in retinal cells. Consistent with previous studies (73), Wnt3a-conditioned medium (WCM) elevated the levels of phosphorylated LRP6 (pLRP6) by 2-fold in ARPE19 cells, a human retinal pigment epithelial cell line, compared to the L-conditioned medium (LCM) control, demonstrating activation of the Wnt pathway by Wnt3a (Fig. 3A). ARPE19 cells express LRP6 at detectable levels, and WCM increased pLRP6 levels in these cells (Fig. 3A). PEDF inhibited the Wnt3a-induced increase of pLRP6 in a concentration-dependent manner (Fig. 3A).

Fig. 3.

The inhibitory effect of PEDF on Wnt3a-induced Wnt pathway activation. ARPE19 cells were incubated with LCM or WCM plus various concentrations of PEDF. (A) After a 2-h incubation with PEDF, pLRP6 and total LRP6 levels were measured by Western blot analysis, quantified by densitometry, and normalized by β-actin levels (mean ± standard deviation; n = 3). *, P < 0.05; **, P < 0.01. (B) Blockage of Wnt3a-induced nuclear translocation of β-catenin by PEDF. ARPE19 cells were treated with LCM or WCM plus 160 nM PEDF for 4 h. β-Catenin was immunostained red (frames a to c), and the nuclei were counterstained blue (frames d to f) and examined using confocal microscopy. Superimposed images of frames a to c and d to f are shown in frames g to i. (C) The nuclear proteins were isolated from ARPE19 cells treated with LCM, WCM, or WCM plus different concentrations of PEDF or DKK1 for 4 h. The nuclear β-catenin levels were measured by Western blot analysis and normalized by lamin B. (D) ARPE19 cells were transfected with the TopFlash vector and then subjected to the same treatments as that described for panel C for 12 h. The luciferase reporter activity was measured with a luminometer (mean ± standard deviation; n = 3). (E) After the same treatment as in described for panel C for 16 h, secreted VEGF levels in the medium were measured by ELISA (mean ± standard deviation; n = 4).

We also evaluated the effect of PEDF on the nuclear translocation of β-catenin induced by Wnt3a. As shown by confocal microscopy, WCM induced an apparent increase of β-catenin in the nuclei of ARPE19 cells (Fig. 3B). PEDF effectively blocked the Wnt3a-induced nuclear translocation of β-catenin (Fig. 3B). To further confirm the effect of PEDF on the Wnt3a-induced β-catenin translocation to the nucleus, we isolated nuclear proteins from the cells treated with LCM, WCM, or WCM plus PEDF and measured the nuclear β-catenin levels. Consistent with the immunocytochemistry results, PEDF decreased nuclear β-catenin levels in the cells treated with Wnt3a (Fig. 3C).

To further characterize the effect of PEDF on Wnt3a-induced Wnt signaling, the transcriptional activity of TCF/β-catenin was also determined using the TopFlash assay, which measures the activity of the luciferase reporter enhanced by the TCF/β-catenin (14, 51). ARPE19 cells were transfected with the TopFlash construct and then exposed to LCM or WCM in the presence of 0, 40 and 80 nM PEDF for 12 h; 10 and 20 nM DKK1 were used as positive controls. The TopFlash assay revealed that the Wnt3a-induced luciferase activity was attenuated by PEDF and by DKK1 in a concentration-dependent manner (Fig. 3D). Similarly, PEDF also inhibited Wnt signaling in other retinal cells. PEDF attenuated the Wnt3a-induced LRP6 phosphorylation and β-catenin nuclear translocation in human retinal endothelial cells and pericytes (data not shown). Furthermore, PEDF inhibited TopFlash activity in human retinal Müller cells and retinal endothelial cells treated with Wnt3a (Fig. 4).

Fig. 4.

PEDF inhibits Wnt3a-induced Wnt signaling in retinal cells. Human retinal endothelial cells (A) and human retinal Müller cells (B) were cotransfected with TopFlash and pRL-TK vectors and then exposed to LCM or WCM. Various concentrations of recombinant PEDF and DKK1 were added to the concentrations as indicated. The activity of the Wnt pathway was measured by TopFlash assay. Firefly luciferase activity was measured and normalized to Renilla luciferase activity (mean ± standard deviation; n = 3). *, P < 0.05; **, P < 0.01.

To correlate the inhibitory effect of PEDF on Wnt signaling with its antiangiogenic activity, we measured the secretion of VEGF, a major proangiogenic factor responsible for retinal NV. ELISA showed that VEGF secretion was induced by WCM exposure, while PEDF and DKK1 both attenuated the overexpression of VEGF induced by Wnt3a (Fig. 3E).

Interestingly, PEDF also decreased total LRP6 levels in a concentration-dependent manner (Fig. 3A). We hypothesize that PEDF binding may induce the endocytosis of membrane-associated LRP6 for degradation. To examine this hypothesis, we measured total LRP6 levels at the cell surface using biotinylation of the cell surface proteins at different incubation time points with Wnt3a or Wnt3a plus PEDF. The results showed that PEDF suppressed the Wnt3a-induced increase of total LRP6 levels and LRP6 levels in the membrane (Fig. 5).

Fig. 5.

PEDF decreases LRP6 levels on the cell surface. (A) ARPE19 cells were treated with WCM in the absence or presence of PEDF for the indicated period of time. Cell lysates were analyzed by Western blotting using anti-pLRP and anti-total LRP6 antibodies. (B) After the treatment with WCM in the absence or presence of PEDF for 2 h, ARPE19 cell surface proteins were biotinylated, extracted, and precipitated with avidin-agarose beads. Precipitates were analyzed for the presence of LRP6 protein by Western blotting using antibodies for LRP6, biotin (multiple surface proteins), and cadherin. The results showed that PEDF decreased LRP6 levels on the membrane, suggesting that PEDF may enhance LRP6 endocytosis. IB, immunoblotting.

To further evaluate the effect of PEDF on regulation of other Wnt target genes, we used quantitative real-time RT-PCR (qRT-PCR) to measure the mRNA levels of Axin2, Cyclin D1, c-Myc, and CTGF in ARPE19 cells treated with PEDF or DKK1 in the presence or absence of Wnt3a stimulation. qRT-PCR showed that both PEDF and DKK1 suppressed the Wnt3a-induced overexpression of Axin2, Cyclin D1, c-Myc, and CTGF mRNA levels (Fig. 6). Moreover, to determine the specificity of PEDF inhibition on the Wnt signaling pathway, we evaluated the effect of PEDF on the activation of other signaling pathways including EGF receptor (EGFR) and G protein-coupled receptor (GPCR); the results showed that PEDF has no effect on these signaling pathways (data not shown).

Fig. 6.

PEDF downregulates the expression of other Wnt3a target genes. ARPE19 cells were cultured in the presence of LCM or WCM plus various concentrations of PEDF or DKK1 for 16 h. The mRNA levels of Axin2 (A), Cyclin D1 (B), c-Myc (C), and CTGF (D) were measured by real-time PCR and are presented as relative values to 50% WCM alone (mean ± standard deviation; n = 5). PEDF and DKK1 both attenuated the Wnt3a-induced Cyclin D1, c-Myc, and CTGF expression. *, P < 0.05; **, P < 0.01.

Enhanced activation of the Wnt signaling pathway by knockdown of PEDF expression.

To determine whether silencing of PEDF expression by small interfering RNA (siRNA) could enhance the activation of Wnt signaling, we evaluated the effects of the PEDF siRNA on the nuclear β-catenin levels, TopFlash activity, and VEGF expression in the absence and presence of WCM. As shown by Western blot analysis, the PEDF siRNA decreased PEDF levels by 65% in ARPE19 cells (Fig. 7A), while the nuclear β-catenin levels were increased in the cells with or without WCM (Fig. 7A). Furthermore, the PEDF siRNA significantly increased the TopFlash activity in the cells (Fig. 7B). ELISA showed that VEGF secretion was also significantly enhanced by the PEDF siRNA in both the absence and presence of WCM (Fig. 7C). Taken together, the results demonstrated that knockdown of endogenous PEDF is sufficient to enhance Wnt signaling.

Fig. 7.

Effect of PEDF siRNA on Wnt signaling. ARPE19 cells were transfected with either an siRNA specific for PEDF (siPEDF) or a control siRNA with a scrambled sequence (siScramble). The cells were cultured for 24 h and then treated with either LCM or WCM for another 16 h. (A) The nuclear β-catenin and cellular PEDF levels were measured by Western blot analysis. The blots are representative of three independent experiments. (B) The cells transfected with siPEDF or siScramble were transfected with the TopFlash plasmid and then exposed to either LCM or WCM for another 16 h. TopFlash activity was measured using a luminometer (mean ± standard deviation; n = 6). (C) VEGF levels in the culture medium from cells transfected with siPEDF or siScramble were measured by ELISA after a 16-h treatment with WCM and expressed as a percentage of that of untreated cells (mean ± standard deviation; n = 5). *, P < 0.05; **, P < 0.01.

PEDF regulates Wnt signaling at the receptor level.

To identify the molecular target through which PEDF inhibits the Wnt pathway, we evaluated the effects of PEDF on Wnt signaling activated by either LiCl, an inhibitor of GSK3β, or a constitutively active mutant of β-catenin. Western blot analysis showed that PEDF did not block the increase of the nuclear β-catenin levels or the TopFlash activity induced by 20 mM LiCl (Fig. 8A and B). Similarly, PEDF did not inhibit the VEGF overexpression induced by LiCl (Fig. 8C). Furthermore, knockdown of GSK3β in ARPE19 cells by a GSK3β-specific siRNA resulted in increased nuclear β-catenin levels, TopFlash activities, and VEGF expression, and these increases were not attenuated by PEDF at various concentrations (data not shown). Likewise, expression of a constitutively active mutant of β-catenin in which a phosphorylation site, Ser37, was mutated to an Ala residue (S37A) resulted in substantially higher nuclear levels of β-catenin and TopFlash activities; these changes were not blocked by PEDF (Fig. 8D and E). These observations suggest that PEDF does not block Wnt signaling intracellularly.

Fig. 8.

Lack of an inhibitory effect of PEDF on Wnt signaling activated intracellularly. (A) ARPE19 cells were treated with 20 mM LiCl plus various concentrations of PEDF for 6 h. Nuclear β-catenin levels were measured by Western blot analysis. (B) ARPE19 cells were transfected with the TopFlash vector overnight and then exposed to 20 mM LiCl plus PEDF for 12 h with 20 mM NaCl as control. TopFlash activity was measured using a luminometer (mean ± standard deviation; n = 5). (C) VEGF levels secreted from the cells treated with LiCl and PEDF were quantified by ELISA (mean ± standard deviation; n = 4). (D and E) The cells infected with either an adenovirus expressing a constitutively active mutant of β-catenin (Ad-β-cat-S37A) at an MOI of 10 or the same titer of Ad-β-Gal were treated with different concentrations of PEDF for 48 h. Total β-catenin levels were measured by Western blot analysis (D); secreted VEGF was quantified by ELISA (E) (mean ± standard deviation; n = 3).

PEDF binds to LRP6 with high affinity.

To reveal potential binding of PEDF to LRP6, we first labeled PEDF with FITC (FITC-PEDF) and BSA with FITC (FITC-BSA) as a control. To show the activity of FITC-PEDF on Wnt signaling, we measured the effect of FITC-PEDF on TCF/β-catenin transcriptional activity using a TopFlash assay with different concentrations of FITC-PEDF in the presence of WCM. The results showed that FITC-PEDF has activities similar to unlabeled PEDF (data not shown). FITC-PEDF and FITC-BSA were separately incubated with ARPE19 cells for 2 h. After thorough washes, the cells were immunostained with an antibody for LRP6, and the nuclei were counterstained with DAPI. As viewed by confocal microscopy, FITC-PEDF signals were detected on the cells incubated with FITC-PEDF. Under the same binding conditions, FITC-BSA signal was not detected in the cells (Fig. 9A). In contrast, under the same conditions, knockdown of LRP6 by an LRP6-specific siRNA (Fig. 9B) decreased PEDF-FITC signal on the cells, suggesting that LRP6 is essential for binding of PEDF to the cells (Fig. 9).

Fig. 9.

Binding of PEDF on the cells expressing LRP6. ARPE19 cells and those transfected with a LRP6 siRNA (siLRP6) or scrambled siRNA were incubated with FITC-PEDF or FITC-BSA for 1 h following thorough washes. (A) Fluorescence microscopy to visualize FITC-BSA and FITC-PEDF on the cells (green; frames a to d). The cells were immunostained with an anti-LRP6 antibody (red; frames e to h). The nuclei were counterstained with DAPI (frames i to l). The merged images of FITC-PEDF, LRP6, and DAPI signals are shown (frames m to p). (B) ARPE19 cells were transfected with the siRNA for LRP6 or scrambled siRNA. Cell extracts were prepared and immunoblotted with an anti-LRP6 antibody at 2 days posttransfection.

To confirm the specific binding of PEDF with LRP6, various concentrations of FITC-PEDF and FITC-BSA were separately incubated with CHO cells overexpressing LRP6 or VLDLR (the negative control), a membrane protein in the same superfamily as LRP6. After thorough washes, FITC-labeled proteins bound to the cells were quantified by a fluorometer. FITC-PEDF showed a concentration-dependent, saturable binding to the cells expressing LRP6 but showed only basal-level binding to the cells expressing VLDLR (Fig. 10A). Under the same conditions, FITC-BSA showed only basal-level binding to the LRP6-expressing cells (Fig. 10A). Scatchard analysis of the PEDF binding data showed that PEDF bound to LRP6 with a K_d (dissociation constant) of 3.7 nM (Fig. 10B).

Fig. 10.

PEDF binds to LRP6 with high affinity. (A) Saturable binding of FITC-PEDF to CHO cells transfected with a LRP6 or VLDLR expression plasmid. The transfected cells were incubated with various concentrations of FITC-PEDF for 2 h, with FITC-BSA as a control. After thorough washes, the fluorescence intensity on the cells was measured with a fluorometer. (B) Scatchard plot of data shown in panel A. (C) Competition with FITC-PEDF for LRP6 binding by unlabeled PEDF or BSA. The cells expressing LRP6 were incubated with 100 nM FITC-PEDF in the presence of excess amounts of unlabeled PEDF or BSA, as indicated. After thorough washes, FITC-PEDF on the cells was quantified. All of the values are mean ± standard deviations (n = 3). (D and E) PEDF was coprecipitated with LRP6. The PEDF-His was incubated with the cells expressing LRP6-Myc or VLDLR-Myc for 16 h. CRBP-His was used at the same concentration as the negative control. (D) After lysis of the cells and solubilization of the membranes, His-tagged proteins were precipitated with Ni-resin. After thorough washes of the resin, the precipitated proteins were immunoblotted (IB) with antibodies specific for either the Myc or His tag. (E) LRP6-Myc and VLDLR-Myc were immunoprecipitated with an anti-Myc tag antibody, and the precipitated proteins were blotted with antibodies specific for either the His or Myc tag.

To further confirm the PEDF binding specificity, binding of FITC-PEDF to the LRP6-expressing cells was competed with excess amounts of unlabeled PEDF; BSA was used as a control. Binding of FITC-PEDF (100 nM) to the LRP6-expressing cells was competed off by the inclusion of excess amounts of unlabeled PEDF (1 to 10 μM) but not by the same amounts of BSA (Fig. 10C).

To demonstrate the physical interaction of PEDF with LRP6, coprecipitation assays were performed. Conditioned medium (CM) containing PEDF tagged with six histidines (PEDF-His) was incubated with cells expressing either LRP6 with a Myc tag (LRP6-Myc) or VLDLR-Myc. CM containing another His-tagged protein, cellular retinol-binding protein (CRBP-His), was used as a negative control. After incubation, the cells were lysed, and the membrane was solubilized. The PEDF-His and CRBP-His were precipitated with Ni-resin, and the precipitated proteins were blotted with an antibody for the Myc tag. The blotting results revealed that LRP6-Myc, but not VLDLR-Myc, was coprecipitated with PEDF-His; LRP6-Myc was not coprecipitated with CRBP-His (Fig. 10D). To further confirm the binding, LRP6-Myc and VLDLR-Myc were immunoprecipitated with the anti-Myc antibody. Western blot analysis of the precipitated proteins using an anti-His tag antibody demonstrated that PEDF-His, but not CRBP-His, was coprecipitated with LRP6-Myc (Fig. 10D). In addition, PEDF-His was not precipitated with VLDLR-Myc (Fig. 10E).

LRP5 is also a coreceptor of the Wnt pathway. To determine whether PEDF also binds LRP5, we expressed the extracellular domain of LRP5 with the Myc tag (LRP5N-Myc), which lacks the intracellular and transmembrane domains. The coprecipitation assays showed that PEDF-His was coimmunoprecipitated with LRP5N-Myc (data not shown), suggesting that PEDF binds to LRP5 as well as LRP6.

Binding of PEDF to LRP6 blocks Wnt3a-induced dimerization of LRP6 with the Fz receptor.

CM from cells expressing either the extracellular fragment of LRP6 fused with a Myc tag (LRP6N-Myc) or the extracellular fragment of low-density lipoprotein receptor (LDLR) with a Myc tag (LDLRN-Myc; the negative control) was separately mixed with CM containing Fz8 cysteine-rich domain fused with an IgG fragment (Fz8-CRD-IgG) in the absence or presence of WCM. LRP6N-Myc and LDLRN-Myc were then precipitated with the anti-Myc antibody. Western blot analysis using an antibody for IgG showed that Fz8-CRD-IgG was coprecipitated with LRP6N-Myc, but not with LDLRN-Myc, in the presence of Wnt3a (Fig. 11). This demonstrates the Fz8-LRP6 complex formation induced by Wnt3a. Furthermore, the addition of various concentrations of PEDF to the mixture abolished coprecipitation of Fz8-CRD-IgG with LRP6N in a PEDF concentration-dependent manner. This suggests that PEDF functions as an antagonist of LRP6, blocking the Fz-LRP6 complex formation induced by Wnt3a (Fig. 11).

Fig. 11.

Blockage of the Wnt3a-induced Fz-LRP6 complex formation by PEDF. Fz8-CRD-IgG and LRP6N-Myc were separately expressed in HEK293 cells. The CM from the cells was harvested in the absence or presence of WCM and different concentrations of PEDF-His or CRBP-His (negative-control protein). The CM containing LDLRN-Myc was used as a negative control. After a 4-h incubation at 4°C, the LRP6-Myc and LDLR-Myc in the CM mixtures were precipitated with the anti-Myc antibody. The precipitates were immunoblotted with the anti-Myc antibody, anti-human IgG antibody, and then anti-His tag antibody.

PEDF binds to the E1E2 domain of LRP6.

To map the PEDF-interacting domain of LRP6, several truncation fragments of LRP6 were generated. These included an extracellular fragment composed of the E1 and E2 domains with a His tag (E1E2-His) and another extracellular fragment composed of the E3 and E4 domains with a His tag (E3E4-His) (Fig. 12A). These fragments of LRP6 were expressed in E. coli and purified to homogeneity through a Ni column, and their binding with PEDF was studied using a coprecipitation assay. Precipitation of PEDF-His with Ni-resin pulled down the full-length LRP6-Myc and LRP6N-Myc but not VLDLR-Myc (Fig. 12B). The reverse coprecipitation using an anti-Myc antibody confirmed that PEDF-His was coimmunoprecipitated with LRP6-Myc and LRP6N-Myc but not with LDLRN-Myc and VLDLR-Myc (Fig. 12C). In addition, a coprecipitation assay using an anti-PEDF antibody showed that E1E2-His, but not E3E4-His, was coprecipitated with PEDF (Fig. 12D and E). In a reverse coprecipitation assay, PEDF was coprecipitated with E1E2-His but not with E3E4-His by the Ni-resin. The coprecipitation of PEDF with E1E2 domains was also confirmed using E1E2 and E3E4 peptides expressed in mammalian cells (Fig. 13).

Fig. 12.

Specific binding of PEDF to the E1E2 domain of LRP6. (A) Schematic illustration of LRP6 deletion mutants used in this study. E1-E4, EGF-like domains 1 to 4; TM, transmembrane domain; CP, cytoplasmic domain. (B and C) LRP6-Myc, LRP6N-Myc, VLDLR-Myc, and LDLRN-Myc were separately expressed in HEK293 cells and incubated with PEDF-His or with CRBP-His as a control. PEDF-His and CRBP-His were precipitated by Ni-resin (B). LRP6N-Myc, VLDLRN-Myc, and LDLRN-Myc were precipitated with an anti-Myc antibody (C). The precipitates were immunoblotted with the anti-Myc and anti-His antibodies. (D and E) The E1E2-His, E3E4-His, VLDLRN-Myc, and LDLRN-Myc were incubated with PEDF (no His tag) or BSA and precipitated with an antibody specific for PEDF (D) and Ni-resin (E). The precipitates were then blotted with antibodies for His tag, Myc, PEDF, and BSA.

Fig. 13.

The E1E2 domains of LRP6 are required for binding of PEDF. The 293 cell-conditioned media containing V5-tagged E1E2 (E1E2-V5) and E3E4-V5 were separately incubated with PEDF-His and with BSA as a control. PEDF-His was precipitated with an anti-PEDF antibody. The precipitates were immunoblotted with the antibodies for the V5 tag and His tag. The results showed that E1E2-V5, but not E3E4-Myc, was coprecipitated with PEDF.

DISCUSSION

The canonical Wnt pathway mediates multiple physiological and pathological processes (37, 42, 48, 51, 59, 71). It is believed to play a key role in angiogenesis (23), as Wnt signaling has been shown to regulate angiogenic factors such as VEGF (16, 21). Wnt signaling also regulates gene expression of fibrogenic factors and inflammatory factors (25, 30, 33, 59). The delicate regulation of the Wnt pathway in these processes has not been fully understood. The present study identified PEDF, a member of the serpin family and a potent antiangiogenic factor, as a novel inhibitor of the Wnt pathway. Furthermore, we have shown that PEDF binds to LRP6 with a high affinity and blocks its activation of Wnt signaling.

The antiangiogenic activity of PEDF was first identified by Dawson and colleagues (23). It was shown that in the eye, PEDF is the major angiogenic inhibitor, counter-balancing the proangiogenic activity of VEGF (2, 9, 28, 38). In addition, PEDF has been extensively studied and has been shown to have anti-inflammatory, antioxidant, antifibrogenic, and anticancer activities (4, 26, 76, 80). The receptor(s) or signaling pathway(s) mediating the broad activities of PEDF has not been identified. The present study demonstrates that PEDF blocks Wnt pathway activation induced by ischemia and the Wnt ligand. Furthermore, our results showed that PEDF functions as an antagonist of LRP6 and blocks Fz-LRP6 complex formation induced by Wnt3a. Moreover, PEDF knockout and knockdown enhance the activation of Wnt signaling, suggesting that PEDF is an endogenous inhibitor of the Wnt pathway. These results suggest that blocking the Wnt pathway may represent a novel mechanism by which PEDF exerts its broad effects.

To correlate the inhibitory effect of PEDF on the Wnt pathway with its antiangiogenic activity in vivo, we used OIR, a commonly used model for ischemia-induced retinal NV (18). In the retina of this model, upregulated VEGF expression and suppressed expression of PEDF lead to a disturbed balance between proangiogenic and antiangiogenic factors, resulting in retinal NV (28, 34, 74). In the OIR retina, the Wnt pathway is activated, as shown by nuclear β-catenin accumulation, negatively correlating with the retinal PEDF levels, while PEDF injection suppressed the Wnt pathway activation in the OIR model.

Based on characterization of PEDF-TG mice, we found that overexpression of PEDF does not lead to any severe developmental defects. However, the body weight of the PEDF-TG mice and their growth rates are significantly lower than that of age-matched WT mice with the same genetic background (Fig. 14A). In addition, the size and mass of the PEDF-TG kidney are smaller than those of WT mice (Fig. 14B). These phenotypes suggest that PEDF might play a role in regulating embryo and organ development. It has been reported that PEDF^−/− mice show increased stromal vessels with epithelial cell hyperplasia in the prostate and pancreas, further supporting that PEDF participates in the regulation of the pancreas and prostate development (24). A recent study showed that under normal conditions, retinal vasculature develops normally in PEDF^−/− mice (34), compared to WT mice. However, under hypoxic conditions, PEDF^−/− mice are more sensitive to oxygen-induced vessel obliteration and VEGF overexpression, resulting in more severe retinal NV, than in hypoxia-exposed WT mice (34). These phenotypes of PEDF-TG and PEDF^−/− mice may be ascribed to the interactions of PEDF with Wnt signaling. However, PEDF-TG and PEDF^−/− mice bred normally and were overtly healthy, suggesting that there may be functional overlaps between PEDF and other endogenous Wnt inhibitors during development. Previous studies have identified a number of proteins which regulate Wnt signaling through binding to LRP6, including the DKK family, endostatin, and IGFBP4 (47, 79, 83). Recently, we have shown that another angiogenic factor, SERPINA3K, a member of serine proteinase family which shares 84% sequence homology with PEDF, also binds to LRP6 (78). Considering the complexity of Wnt pathway regulation and its interactions with other pathways, it is possible that the lack of severe developmental defects in PEDF-TG and PEDF^−/− mice may be ascribed to compensation from other Wnt pathway modulators or from other signaling pathways.

Fig. 14.

Phenotypic characteristics of PEDF-TG mice. (A) Body mass of WT and PEDF-TG mice from 1 to 15 weeks of age. (B) Body size of 20-week-old WT and PEDF-TG mice. (C) Size and weight of the kidney from both WT and PEDF-TG mice. All values are the means ± standard error of the means (n = 13 to 15 in each group). The results showed that PEDF-TG mice have a lower growth rate and decreased size and weight of the kidney than the age-matched WT mice.

Overexpression of PEDF in the PEDF-TG retina as well as intravitreal injection of PEDF protein attenuated Wnt pathway activation in the retina of OIR mice. In contrast, the Wnt signaling pathway was activated by silencing PEDF using the PEDF siRNA and in the retina of PEDF^−/− mice. These effects on Wnt signaling correlate with the inhibitory effects of PEDF on the retinal inflammation, vascular leakage, and NV reported previously (34, 60, 80). Taken together, these observations suggest that blockage of the Wnt pathway is responsible, at least in part, for both the anti-inflammatory and antiangiogenic activities of PEDF.

Our recent study showed that VLDLR is a negative regulator of the Wnt pathway (16). VLDLR gene knockout resulted in activation of the Wnt pathway in the retina and retinal pigment epithelium (RPE) and, consequently, subretinal NV (16). Here, our results also showed that administration of PEDF protein suppressed Wnt pathway overactivation in Vldlr^−/− mice. These findings provide further support for the LRP6-antagonizing activity of PEDF.

To investigate if the inhibition of Wnt signaling by PEDF is via a direct interaction with the Wnt pathway, we evaluated the inhibitory effects of PEDF on the Wnt3a-activated Wnt pathway. In cultured cells, PEDF inhibited the phosphorylation of LRP6 and decreased total LRP6 and nuclear β-catenin levels, suggesting that PEDF decreases total LRP6 levels on the plasma membrane. It has been shown that binding of DKK1 to LRP6 induces degradation of LRP6 via endocytosis (41, 77). Furthermore, PEDF attenuated the transcriptional activity of TCF/β-catenin, as shown by the luciferase reporter activity assay. Silencing PEDF expression by siRNA enhanced the activation of the Wnt signaling pathway, further supporting the role of PEDF as an endogenous inhibitor of the Wnt pathway. In contrast, PEDF did not block the intracellular cascade of the Wnt signaling pathway, as seen with the GSK3β inhibitor, LiCl, and a constitutively active mutant of β-catenin. These results suggest that PEDF functions as a Wnt pathway inhibitor, acting on a cell surface receptor. This conclusion was further supported by the specific binding of FITC-PEDF to LRP6, a coreceptor of the canonical Wnt pathway. In addition, the coprecipitation of PEDF with either full-length LRP6 or with the extracellular fragment of LRP6 further confirmed a physical interaction between PEDF and LRP6.

In the canonical Wnt pathway, some Wnt ligands bind to the Fz receptor and to LRP6, inducing the dimerization and formation of a receptor complex between Fz and LRP6, an essential step in activation of the downstream signaling cascade. DKK1, a specific inhibitor of the Wnt pathway, binds to LRP6 in the E3E4 domain and blocks the dimerization between LRP6 and the Fz receptor (41, 50, 58). Another research group showed that DKK1 binds to both the E1E2 and E3E4 fragments of LRP6 and competes with Wnt3a (13). This disparity suggests that the mechanism for DKK1 activity remains uncertain. To determine whether PEDF competes with Wnt ligand for binding to LRP6, we performed an ELISA-based in vitro PEDF binding assay and found that Wnt3a does not compete with PEDF for binding to LRP6 (data not shown). This result suggests that the inhibitory effect of PEDF is not through blocking Wnt ligand binding. Our coprecipitation assay showed that PEDF binds to LRP6 at the E1E2 domain. Further, a coimmunoprecipitation assay showed that binding of PEDF to LRP6 hinders the Wnt ligand-induced dimerization of LRP6 and Fz receptor and, thus, blocks the activation of the Wnt pathway (Fig. 11). These observations suggest that PEDF may inhibit Wnt signaling through a mechanism by inducing structural change of LRP6 upon binding to the E1E2 region of LRP6. Interestingly, we also found that the N-terminal region, but not the C terminus, of PEDF binds to LRP6N and inhibits Wnt pathway activation (Fig. 15). This finding is consistent with previous structure-function studies showing that the N-terminal region of PEDF is responsible for its antiangiogenic activity. These results indicate that the inhibitory effect on Wnt signaling and the antiangiogenic activity reside in the same domain of PEDF (26). Another possible mechanism is that binding of PEDF to LRP6 may induce the endocytosis and subsequent degradation of membrane-bound LRP6. Biotinylation assay showed that PEDF decreased cell surface LRP6 levels, supporting this possible mechanism (Fig. 5). However, the molecular mechanism underlying the interaction of PEDF with LRP6 remains to be further investigated.

Fig. 15.

The N-terminal region of PEDF inhibits the canonical Wnt pathway. (A) Constructs of PEDF deletion mutants used in this study. SS, signal sequence; N, N terminus; C, C terminus; PEDF-CT, PEDF C-terminal peptide. FL, full-length. Residue positions are given in parentheses. (B) ARPE19 cells were incubated with WCM or LCM and then treated with PEDF deletion mutants. Nuclear β-catenin levels were measured with Western blot analysis. (C) ARPE19 cells were transfected with the TopFlash vector and then subjected to the same treatments as described for panel B for 12 h. The luciferase activity was measured with a luminometer (mean ± standard deviation; n = 3). *, P < 0.05; **, P < 0.01. (D) LRP6N-Myc and LDLRN-Myc were separately expressed in HEK293 cells and incubated with the His-tagged PEDF deletion mutants. The His-tagged deletion mutants of PEDF were precipitated by Ni-resin, and the precipitates were blotted with antibodies for Myc tag and His tag.

PEDF is produced in most mammalian tissues (8, 15, 19, 56, 68). Among them, the liver and kidney are considered the major sources of PEDF in the circulation (43). In addition, PEDF was found to exist at high levels in the eye in both the RPE and interphotoreceptor space (66, 75). For instance, PEDF levels are 4.6 to 6.1 μg/ml in normal human serum, 1.5 to 16.4 μg/ml in the vitreous fluid, and 7.2 μg/ml in the interphotoreceptor space (6, 10, 35, 64, 69). The present study showed that PEDF binds to LRP6 with a K_d of 3.7 nM (0.185 μg/ml), which is substantially lower than its physiological concentrations in the serum and vitreous and interphotoreceptor space (6). Although PEDF has a lower binding affinity to LRP6 than to DKK1 (K_d of 0.34 to 0.39 nM) (5, 41), physiological levels of PEDF in the serum are substantially higher than the level of DKK1, which could compensate for its relatively lower affinity. These results suggest that PEDF is likely to function as an endogenous antagonist of LRP6. The potency of PEDF in inhibiting the Wnt pathway is consistent with that for its antiangiogenic and anti-inflammatory effects (23, 80). Interestingly, PEDF has been reported to bind to some cell surface molecules, such as heparin sulfate, collagens, and phospholipase A₂ξ (PLA₂ξ) (49). The interactions of these molecules with LRP6 are unknown. However, it is possible that binding to some of these cell surface molecules, such as heparin sulfate, may potentiate PEDF binding with LRP6.

It has been shown that PEDF levels are decreased in the vitreous and aqueous fluids from patients with DR (9, 10, 27, 54, 64). Several groups have suggested that decreased PEDF levels contribute to inflammation, vascular leakage, and NV in several pathological conditions (27, 64). Our recent studies have shown that the Wnt pathway is overactivated in the retina of human patients with DR and in the retina of streptozotocin (STZ)-induced diabetic rats, Akita mice, and OIR rats (18, 47). Furthermore, blockage of the Wnt pathway using DKK1 ameliorated retinal inflammation, vascular leakage, and NV in these models, suggesting that dysregulation of the Wnt pathway in diabetes plays a pathogenic role in DR (18). As PEDF is a potent inhibitor of the Wnt pathway, decreased PEDF levels in the DR retina and vitreous may contribute, at least in part, to Wnt pathway activation, retinal inflammation, fibrosis, and NV.