Pigment epithelium–derived factor restoration increases bone mass and improves bone plasticity in a model of osteogenesis imperfecta type VI via Wnt3a blockade

Abstract

Null mutations in for pigment epithelium–derived factor (PEDF), the protein product of the SERPINF1 gene, are the cause of osteogenesis imperfecta (OI) type VI. The PEDF-knockout (KO) mouse captures crucial elements of the human disease, including diminished bone mineralization and propensity to fracture. Our group and others have demonstrated that PEDF directs human mesenchymal stem cell (hMSC) commitment to the osteoblast lineage and modulates Wnt/β-catenin signaling, a major regulator of bone development; however, the ability of PEDF to restore bone mass in a mouse model of OI type VI has not been determined. In this study, PEDF delivery increased trabecular bone volume/total volume by 52% in 6-mo-old PEDF-KO mice but not in wild-type mice. In young (19-d-old) PEDF-KO mice, PEDF restoration increased bone volume fraction by 35% and enhanced biomechanical parameters of bone plasticity. A Wnt–green fluorescent protein reporter demonstrated dynamic changes in Wnt/β-catenin signaling characterized by early activation and marked suppression during terminal differentiation of hMSCs. Continuous Wnt3a exposure impeded mineralization of hMSCs, whereas the combination of Wnt3a and PEDF potentiated mineralization. Interrogation of the PEDF sequence identified a conserved motif found in other Wnt modulators, such as the dickkopf proteins. Mutation of a single amino acid on a 34-mer PEDF peptide increased mineralization of hMSC cultures compared with the native peptide sequence. These results indicate that PEDF counters Wnt signaling to allow for osteoblast differentiation and provides a mechanistic insight into how the PEDF null state results in OI type VI.—Belinsky, G. S., Sreekumar, B., Andrejecsk, J. W., Saltzman, W. M., Gong, J., Herzog, R. I., Lin, S., Horsley, V., Carpenter, T. O., Chung, C. Pigment epithelium–derived factor restoration increases bone mass and improves bone plasticity in a model of osteogenesis imperfecta type VI via Wnt3a blockade.

Osteogenesis imperfecta (OI) are a heritable collection of bone diseases known by the clinical moniker, brittle bone disease, and whose diverse genetic causes are increasingly being identified (1, 2). OI type VI is an autosomal recessive disease that is characterized by decreased bone mineralization and early fractures (3). Exome sequencing has identified null mutations in the pigment epithelium–derived factor (PEDF) gene, Serpinf1, as the cause of OI type VI, with the pathologic hallmark of decreased mineralization in the setting of normal collagen (4–9). Another OI subtype, type V, that is characterized by a specific S40L mutation in BRIL protein, results in nearly undetectable PEDF levels and a bone biopsy characteristic of OI type VI, which suggests a comprehensive role for PEDF in bone development (10).

PEDF is a 50-kDa secreted protein that circulates in humans at concentrations of ∼100 nM but whose expression is nearly absent in adult bone (11, 12). The PEDF-knockout (KO) mouse recapitulates human OI type VI with decreased bone mineralization, reduced trabecular bone volume, and susceptibility to fracture upon biomechanical testing (13). Whether systemic reconstitution of PEDF could enhance the bone phenotype in the murine model of OI type VI has not been demonstrated. Furthermore, our group and others have previously reported that PEDF induces human and murine PEDF-null mesenchymal stem cells to the osteoblast lineage by modulating Wnt/β-catenin signaling, a major factor in bone development and homeostasis (14, 15). RNA interference of endogenous PEDF in human mesenchymal stem cells (hMSCs) and PEDF treatment led to suppression of sclerostin and increased levels of β-catenin (16). The precise temporal role of PEDF in modulating Wnt/β-catenin signaling, however, remains unclear.

Current treatments to enhance bone strength in patients with OI primarily involve bisphosphonate administration. This approach is less successful for patients with OI type VI as reductions in fracture incidence are less improved than for patients with other OI subtypes (17). In this study, PEDF restoration was sufficient to increase trabecular and cortical bone mass as well as to improve parameters of bone plasticity in PEDF-KO mice. Furthermore, PEDF counters the effects of canonical Wnt3a ligands in the terminal phase of osteoblast maturation. Unopposed Wnt3a impairs bone mineralization, but coadministration of PEDF or a PEDF peptide enhanced osteoblast mineralization, which was consistent with a Wnt antagonist function that has been described for PEDF in other organ systems (18–20). A motif found on other Wnt modulators that interacts with a specific domain on the Wnt receptor LRP6 is also found on PEDF (21). Mutation of a single amino acid within this motif enhanced LRP6 blockade and conferred gain-of-function properties, thereby highlighting the role of PEDF in Wnt/β-catenin signaling.

MATERIALS AND METHODS

Animals

PEDF-KO mice have been described in Doll, et al. (22). PEDF-KO mice were bred with C57BL/6 wild-type (WT) mice to generate heterozygous breeding pairs. Mice were backcrossed for >10 generations to generate KO and WT breeding pairs. Mice were maintained in normal, specific pathogen–free conditions on 12-h light/dark cycles and fed with a standard mouse chow diet ad lib. All animal experiments were done in accordance with the Yale University and Veterans Affairs Connecticut Health Care System Institutional Animal Use and Care Committees.

Materials

Sterile alginate was purchased from NovaMatrix (Sandvika, Norway). Hydroxypropylmethylcellulose (HPMC) was obtained from Sigma-Aldrich (St. Louis, MO, USA). Calcium chloride was from J. T. Baker (Center Valley, PA, USA). Iso-octane, microBCA assay, and cell culture reagents were from Thermo Fisher Scientific Life Sciences (Waltham, MA, USA), Span 80 from Fluka (Buchs, Germany), Tween 80 from Riedel-de Haen (Seelze, Germany), and 2-propanol from American Bio (Natick, MA, USA).

PEDF, PEDF peptides, synthesis of PEDF-containing microspheres, and PEDF restoration

Full-length recombinant human PEDF was generated in human embryonic kidney cells (23). A 34-mer fragment of PEDF (44–77 aa) and its mutated versions were commercially obtained (sequence: DPFFKVPVNKLAAAVSNFGYDLYRVRSSTSPTTN, N-acetylation and C-amidation; NeoBiolab, Cambridge, MA, USA). PEDF integrity was confirmed by SimplyBlue staining (Thermo Fisher Scientific Life Sciences) and was then dialyzed in PBS. PEDF-containing microspheres were prepared on the basis of the method described by Zheng et al. (24, 25), with several modifications. In brief, sterile alginate (2%) and HPMC (0.2%) were dissolved in ultrapure H₂O, followed by direct dissolution of PEDF for a theoretical maximum of 0.2 µg PEDF/mg alginate. Iso-octane +5% (v/v) Span 80 was homogenized at 17,500 rpm, and the alginate/HPMC/PEDF solution was added dropwise. Then 30% (v/v) aqueous Tween 80 was added dropwise, and the emulsion was mixed for 3 min. An aqueous solution of calcium chloride (100 mM) was added at a rate of 4 ml/min. After mixing, 2-propanol was added and the particles were allowed to cure for 3 min. Particles and supernatant were centrifuged at 4000 rpm for 1 min. Supernatant was then removed and particles were washed 2 times in 2-propanol and air dried. Particles were resuspended in ultrapure H₂O and lyophilized. PEDF loading data was analyzed with the microBCA assay according to the manufacturer protocol. Temporal release of PEDF microspheres was assessed by ELISA (BioProducts MD, Middletown, MD, USA) as shown in Supplemental Fig. 2.

Intraperitoneal injections were administered to 6-mo-old male WT and PEDF-KO mice (n = 4/group) as previously reported (19). PEDF (25 μg/kg body weight) was injected intraperitoneally on alternate days for a period of 4 wk. Another set of young (19-d-old) male mice were administered a 1-time intraperitoneal injection of PEDF-containing microspheres (150 ng/g body weight; n = 10–11/group) and bones collected after 3 wk. Bones in which the growth plate was detached were excluded from the analysis. Tibias and femurs from mice were processed at the Yale Core Center for Musculoskeletal Disorders for bone micro–computed tomography (µCT) analysis and biomechanical testing.

Bone µCT and histologic analysis

A ScanCo µCT 35 scanner (Yale Center Core for Musculoskeletal Disorders) was used to assess the distal femur for trabecular and midshaft for cortical bone morphology from WT and PEDF-KO mice in a blinded manner. Axial, sagittal, and coronal images were obtained at standardized sites, and measures of trabecular, total bone volumes, and other parameters of bone density were obtained (26). For histology, femurs and tibias of day-old mice were dissected, cleaned, and fixed in 70% ethanol, then further dehydrated via graded ethanols, cleared in toluene, and embedded in methyl methacrylate. After polymerization, methyl methacrylate blocks were removed from the mold, cut to size, sanded, and polished on a Buehler MetaServ (Buehler, Lake Bluff, IL, USA). Longitudinal sections of 5-μ thickness were cut by using a Reichert-Jung RM 2165 microtome with a d-profile tungsten carbide knife (Reichert Technologies, Buffalo, NY, USA), mounted on charged slides, and stained with Toluidine Blue O (pH 3.7).

Bone mechanical analysis

Femurs were loaded to failure in 4-point bending. All whole-bone tests were conducted by loading the femur in the posterior to anterior direction such that the anterior quadrant was subjected to tensile loads. Widths of the lower and upper supports of the 4-point bending apparatus were 7 mm and 3 mm, respectively. Tests were conducted with a deflection rate of 0.05 mm/s by using a servo-hydraulic testing machine (Instron model 8874; Instron, Norwood, MA, USA). Load and midspan deflection were acquired directly at a sampling frequency of 200 Hz. Load-deflection curves were analyzed for stiffness, maximum load, and work to fracture. Yield is defined as a 10% reduction in the secant stiffness (load range normalized for deflection range) relative to initial tangent stiffness. Postyield deflection, defined as the deflection at failure minus the deflection at yield, was also measured. Femurs were tested at room temperature and kept moist with PBS.

Bone protein extraction of PEDF-injected mice

To ascertain the effects of PEDF on Wnt signaling in vivo, 2-mo-old PEDF-KO mice were injected intraperitoneally with 100 ng/g of PEDF protein or PBS control. One day later, femurs and tibias were removed from euthanized mice, immersed in ice-cold PBS, and bones were snap-frozen in liquid nitrogen. Bones were ground in liquid nitrogen with a mortar and pestle, and then extracted with commercial lysis buffer (mammalian protein extraction reagent; Pierce, Rockford, IL, USA) that contained protease and phosphatase inhibitors. Protein concentrations were measured by using a Bio-Rad protein assay (500-0006; Hercules, CA, USA) with BSA standards, and 50 µg protein was loaded per lane, then run on a 4–20% gradient gel.

Mineralization in vitro

hMSCs [PCS-500-011; American Type Culture Collection (ATCC), Manassas, VA, USA] were propagated in MSC basal medium that was supplemented with MSC growth kit (ATCC). Cells were CD29-, CD44-, CD105-, and CD166-positive and were negative for CD31, CD34, and CD45. Directly after thawing, passage 2 cells were seeded at 3 × 10⁴ cells per well into Falcon 24-well plates in MSC basal medium. Two days later (d 0), medium was switched to Osteocyte Differentiation Tool (PCS-500-052 with 0.5× penicillin/streptomycin/fungizone; ATCC) with the indicated treatments. Medium was changed every 3–4 d. For late PEDF or PEDF peptide groups, 300 ng/ml PEDF or 100 nM peptide was added from d 13 to 21. Interrogation of Wnt/β-catenin signaling was done by using passage 2 hMSCs in ATCC MSC basal medium with 2% fetal bovine serum. PEDF (500 ng/ml) was added for 6 h.

Staining

To assess mineralization with Alizarin red S staining, cells were fixed for 10 min with 10% formaldehyde that was added directly to the medium. After washing in PBS, cells were incubated with cold 70% ethanol for 30 min. Cells were then stained with 2% Alizarin red S (pH 4.2) for 15 min, and wells were rinsed with water 6 times. Pictures were captured (two 0.5 × 0.5 cm pictures/well) on a dissecting scope and red staining was quantitated by using ImageJ (National Institutes of Health, Bethesda, MD, USA).

Lentiviral transduction of hMSC with Wnt–green fluorescent protein reporter

The Wnt–green fluorescent protein (GFP) reporter plasmid 7TGP was deposited from the laboratory of Roel Nusse (plasmid no. 24305; Addgene, Cambridge, MA, USA), which contains the 7xTcf-eGFP/SV40-PuroR insert. Lentiviral particle viral supernatants were used to transduce hMSCs seeded at 4 × 10⁴ cells/well in the presence of 6 µg/ml polybrene (Millipore, Billerica, MA, USA). Medium was changed after 17 h, and puromycin (2 µg/ml) was added 3 d later. No appreciable cell death was observed in transduced cells, whereas all polybrene-treated, nontransduced cells died after 3 d in puromycin. For these experiments, cells were differentiated in MEM-α without phenol red supplemented with 2% fetal bovine serum, 50 µg/ml ascorbic acid 2 phosphate, 10 nM dexamethasone, and 2.5 mM β-glycerol-phosphate.

RNA analysis and quantitative PCR

RNA was isolated and processed by using Trizol (Thermo Fisher Scientific Life Sciences). Primer probe sets were obtained from a commercial source (Thermo Fisher Scientific Life Sciences), and quantitative RT-PCR was performed on a TaqMan ABI 7500 system (Thermo Fisher Scientific Life Sciences) as previously described (27). Target gene expression was normalized against 18S ribosomal RNA by using the ΔΔC_t method.

Immunoblotting

Immunoblotting was performed as described in Schmitz et al. (27). Proteins were separated by SDS-PAGE on 4–15% gradient gels (Bio-Rad). Antibodies were obtained against nonphosphorylated (active) β-catenin and total β-catenin, (Cell Signaling Technologies, Danvers, MA, USA). Glyceraldehyde 3-phosphate dehydrogenase (Santa Cruz Biotechnology, Santa Cruz, CA, USA) was used as loading control. For PEDF Western blots, 10 µl of conditioned medium was loaded per lane and probed with anti-PEDF (Millipore).

Statistical analysis

Results were assessed by using Student’s t test to compare 2 groups or by 1-way ANOVA with Bonferroni post hoc test for comparisons between groups and expressed as means ± sd. A value of P < 0.05 was considered significant.

RESULTS

PEDF increases trabecular bone mass in adult PEDF-KO mice

An initial study of PEDF injection into 6-mo-old WT and PEDF-KO mice (n = 4/group) determined that PEDF could increase trabecular bone mass in PEDF-KO but not in WT mice. PEDF restoration increased trabecular bone volume/total volume (BV/TV) ratio by 52% in PEDF-KO mice but did not significantly change trabecular bone volume in WT mice (Fig. 1). Trabecular thickness was significantly increased (Fig. 1B), and other parameters, such as apparent density (P = 0.01) and bone surface (P = 0.07), were significantly increased in KO mice that were administered PEDF but not in WT mice (Supplemental Fig. 1). Thus, PEDF increased bone mass in adult PEDF-KO but not in WT mice.

Figure 1.

PEDF increases trabecular bone volume/total volume (BV/TV), the bone volume fraction, in 6-mo-old mouse femurs. A) μCT and Toluidine blue images of adult mouse femurs after treatment with vehicle or PEDF. WT or PEDF-KO as indicated. B) BV/TV and trabecular thickness parameters for adult PEDF-KO mice and WT littermates. Values are expressed as means ± sd; n = 4 mice/group. Original magnification, ×4.

PEDF increases bone mass in young PEDF-KO mice

Because a trophic effect on bones was seen in mature PEDF-KO mice and we previously corrected a metabolic effect in PEDF-KO mice by using PEDF-containing microspheres (19), we next tested whether PEDF restoration could increase bone mass in young PEDF-KO mice. PEDF-KO mice (19 d old) were treated with vehicle or PEDF-containing microspheres and euthanized 3 wk after injection. μCT images from PEDF-treated mice demonstrated increases in trabecular bone volume (Fig. 2A). This was reflected by a 35% increase in trabecular bone volume fraction, BV/TV, in PEDF-treated mice compared with controls (Fig. 2B). There was a significantly greater trabecular number (20%) and corresponding 16% decrease in trabecular separation after PEDF restoration without a significant change in trabecular thickness (Supplemental Fig. 3). Similarly, connectivity density increased after PEDF restoration (Fig. 2B). These results show that PEDF restoration increases trabecular bone volumes and density in the murine model of OI type VI.

Figure 2.

PEDF increases bone mass in young (19-d-old) mice. A) Representative µCT images of femurs from vehicle- and PEDF-treated mice where PEDF restoration enhanced trabecular bone volume. B) Quantification of trabecular bone volumes after PEDF. C) Increased cortical BV/TV and cortical thickness after PEDF treatment. Values are expressed as means ± sd; n = 9 mice/group.

PEDF administration also enhanced cortical bone volume parameters. Cortical BV/TV and thickness were increased by PEDF (Fig. 2C). Total area of cortical bone as reflected in the cross-sectional area was not different between the 2 groups (Supplemental Fig. 3). PEDF treatment significantly increased the cortical area fraction, which reflected enhanced cortical thickness with PEDF, but did not affect endosteal or periosteal radii (Supplemental Fig. 3). Significant differences in bone tissue density and bending moment of inertia were not found, although for both of these end points, PEDF-treated mice had a higher mean value than did controls (Supplemental Fig. 3). Thus, PEDF modestly enhanced several cortical bone volume parameters in conjunction with significant effects seen in trabecular bone.

PEDF treatment improves biomechanical parameters of femurs in KO mice

Biomechanical testing of femurs from mice that were treated with PEDF demonstrated improvements in plasticity compared with vehicle-treated mice. Femur length was not different between the 2 groups (Fig. 3A). PEDF-treated animals showed a 50% increase in stiffness (slope of the load/displacement; Fig. 3B and Supplemental Fig. 4) without significant differences in maximal load (Fig. 3C). This finding is consistent with a previous report that showed no differences in maximal load between control and PEDF-KO mice, despite the latter having bones that were more prone to break (13). The displacement at yield was not different between groups (Fig. 3E), while the preyield work was significantly reduced compared to vehicle-treated mice (Fig. 3F). Postyield deflection, the amount of bone deformation after the yield point, was >50% higher in PEDF vs. vehicle-treated mice (Fig. 3H). Postyield work was not significantly different between the 2 groups but did correspond to changes in displacement at yield. This increase in bone plasticity was reflected in significantly higher bone displacement at maximum load (Fig. 3G) and displacement at fracture (Fig. 3D). Thus, PEDF treatment in a murine model of OI type VI stiffens the biomechanical properties of bones, whereas, at the same time, permitting a greater degree of bone plasticity upon mechanical forces necessary to cause fracture.

Figure 3.

PEDF enhances the plasticity of femurs in PEDF-KO mice. A) Femur length was not different between groups. B) PEDF increased bone stiffness, the slope of the load vs. displacement. C) Maximum load was not different in PEDF-treated mice. D) Displacement at fracture (FX) increased after PEDF treatment. E) Displacement at yield. F) Preyield work. G) Displacement of femurs upon maximum load. H) Postyield deflection, the deformation of bone after the yield point, which is a measure of bone plasticity, increased after PEDF treatment. I) Postyield work. Values are expressed as means ± sd; n = 3 vehicle-treated mice, n = 5 PEDF-treated mice. *P < 0.05 by Student’s t test.

Endogenous PEDF secretion by hMSCs increases with osteogenesis, and is suppressed by Wnt3a

Previous studies have indicated the possibility of paradoxical effects of diminished mineralization with continuous canonical Wnt ligand exposure of hMSCs, a finding that is seemingly at odds with the paradigm that canonical Wnt signaling drives bone development (28–30). To monitor Wnt signaling over time, hMSCs that were transduced with the Wnt/β-catenin GFP reporter were differentiated by using osteocyte differentiation medium and examined over the course of 21 d. Interrogation of downstream Wnt3a signaling by using a T-cell factor–driven Wnt-GFP reporter showed striking loss of T-cell factor–induced signal over the course of a 21-d hMSC osteogenesis differentiation protocol, which was not observed in control cultures in which ascorbate and dexamethasone were not added (Fig. 4A). This finding is consistent with down-regulation of the canonical Wnt-signaling pathway during terminal osteogenesis. Day 14 quantitative PCR analysis of bone-related genes confirmed increased expression of SP7, RUNX2, ALPL, and BGLAP (Supplemental Fig. 5), which was consistent with the presence of osteoblast-like cells. Endogenously secreted PEDF in conditioned medium from differentiating hMSCs increased over time (Fig. 4B), which suggested that PEDF may play a more important role during the latter stages of osteoblast differentiation. Continuous Wnt3a 50 ng/ml exposure for 21 d to differentiating hMSCs decreased endogenous PEDF levels by 30% in the conditioned medium (Fig. 4C). These results identify temporal changes in canonical Wnt signaling in osteogenic differentiation and characterize endogenous PEDF as a secreted protein during the middle to late stages of this process.

Figure 4.

Endogenous PEDF secretion by hMSCs increases with osteogenesis and is suppressed by Wnt3a. A) hMSC transduced with a Wnt-GFP reporter grown in osteogenic or control medium starting at d 0. Fields imaged at indicated days (original magnification, ×10); GFP⁺ cells counted; n = 5–20 fields per time point. B) Temporal quantitation of endogenously secreted PEDF over 21 d in osteogenic medium, with Coomassie blue–stained gel to show equal loading of conditioned medium. C) Suppression of PEDF secretion in hMSCs at d 21 with continuous Wnt3a exposure. DIC, differential interference contrast. Values are expressed as means ± sd. *P < 0.05 by Student’s t test; **P < 0.0001 compared with control.

Unopposed Wnt3a results in impaired mineralization that is rescued by PEDF and PEDF peptides

To ascertain the effect of PEDF on in vitro mineralization, we examined exogenous PEDF and PEDF peptide reconstitution. Continuous Wnt3a 50 ng/ml exposure for 21 d to differentiating hMSCs reduced mineralization by 40% as measured by Alizarin red staining (Fig. 5A). A conserved motif found on other Wnt inhibitors that bind LRP5/6 has been identified (21). We examined motifs on PEDF that contain a similar conserved motif. Potential homologous regions targeted for modification in the PEDF 34-mer demonstrate conservation across species (Fig. 5B). A lysine to alanine (k→a) substitution at aa 53 in the native PEDF 34-mer was generated to see whether this k→a mutated peptide could have gain-of-function effects relative to the native 34-mer. Because continuous Wnt3a inhibited mineralization (Fig. 5A), PEDF peptides were added to determine if mineralization could be restored. Unmodified PEDF 34-mer and k→a PEDF 34-mer had little effect on mineralization when administered concurrently with Wnt3a; however, native PEDF 34-mer and k→a PEDF 34-mer significantly increased mineralization when added during the last 7 d of the differentiation protocol (Fig. 5C, D). The k→a 34-mer enhanced mineralization by ∼75% compared with Wnt3a-treated cells and by >30% compared with the native peptide (Fig. 5D), which suggests that key residues on PEDF have been identified that interfere with Wnt3a-mediated suppression of mineralization.

Figure 5.

Unopposed Wnt3a results in impaired mineralization that is rescued by PEDF and PEDF peptides. A) Alizarin red S staining of 21-d-old hMSC cultures with continuous Wnt3a (50 ng/ml) exposure vs. controls (ctl). B) PEDF amino acid sequence homology of putative Wnt modulating motif (in boxes). Left box contains lysine that was changed to alanine, k→a. C) Alizarin red S staining of 21-d-old hMSC cultures with continuous Wnt3a 50 ng/ml alone or in combination with PEDF 34-mer 100 nM, or k→a 34-mer. For late groups, PEDF 34-mer was added only during the last 8 d of differentiation protocol. D) Quantitation of Alizarin red S staining in panel C. PEDF 300 ng/ml. Values are expressed as means ± sd; n = 3–4 wells/group. *P < 0.05, **P < 0.01 by Student’s t test compared with Wnt3a group; 1-way ANOVA, P = 0.0022.

PEDF peptides inhibit Wnt/β-catenin signaling

Temporal activation of Wnt/β-catenin signaling is essential for osteoblast differentiation (31). Because Wnt3a-mediated inhibition of mineralization in vitro is paradoxical, we surmised that Wnt signaling modulation by PEDF might be necessary to induce the Wnt downstream transcriptional target, β-catenin. Continuous Wnt3a treatment over 21 d suppressed active (nonphosphorylated) β-catenin and total β-catenin by 60 and 68%, respectively. Reconstitution with the PEDF 34-mer increased active β-catenin and total β-Catenin levels more than that of Wnt3a-treated cells (Fig. 6A, C). These data are consistent with continuous Wnt3a repressing β-catenin to reduce osteoblast differentiation via down-regulation of signal transduction by continual Wnt stimulation (28, 29). To assess the acute effects of PEDF peptides on Wnt signaling compared with other known Wnt antagonists, such as dickkopf (DKK) 1, proteins were extracted 6 h after peptide treatment of hMSCs. Immunoblots demonstrated reduced pLRP6 and active β-catenin, with the k→a PEDF 34-mer having a greater effect than native 34-mer (Fig. 6B, D). DKK1 also inhibited pLRP6 and active β-catenin. When DKK1 and the k→a 34-mer were applied together, no significant additive reduction occurred, which indicated blockade of the same target receptor.

Figure 6.

PEDF peptides inhibit Wnt/β-catenin signaling. A) Western blots on active (nonphosphorylated) β-catenin and total β-catenin after continuous 50 ng/ml Wnt3a treatment or continuous Wnt3a + 100 nM PEDF 34-mer treatment of hMSCs at day 21 of osteogenic differentiation. Duplicate wells done for each treatment. B) Short-term exposure to PEDF peptides or DKK1 antagonizes Wnt signaling in hMSCs. Cells were incubated in 200 nM peptides or 50 ng/ml DKK1 for 6 h, then protein extracts were made and Western blots were performed as indicated. C) Quantitation of bands in panel A. D) Quantitation of bands in panel B. GAPDH, glyceraldehyde 3-phosphate dehydrogenase. Values are expressed as means ± sd; 2 biologic replicates per group.

Acute PEDF treatment inhibits Wnt signaling in murine bone

Immunoblots of adult mouse bone protein extracts 1 d after intraperitoneal PEDF injection revealed inhibitory activity toward LRP6 and active β-catenin (Fig. 7A, B), which was consistent with a Wnt inhibitory effect. Treatment of PEDF-KO mice with PEDF resulted in a 39% reduction of pLRP6, a 62% reduction of total LRP6, and a 60% reduction of active β-catenin. These results are consistent with PEDF acting as an antagonist of the Wnt signaling pathway in mature bone. In early osteoblast specification, we and others have shown a Wnt agonist activity (14, 15). The data in the current work identify a dominant late Wnt inhibitory activity for PEDF. This is represented schematically in Fig. 8.

Figure 7.

PEDF protein restoration in PEDF-KO mice inhibits Wnt signaling. A) Bone proteins were extracted and the indicated immunoblots were performed. B) Quantitation of bands from 2 separate blots; n = 6 bands per treatment. Each lane is a separate tibia + femur. GAPDH, glyceraldehyde 3-phosphate dehydrogenase. Values are expressed as means ± sd. *P < 0.05, **P < 0.001 by Student’s t test.

Figure 8.

Schematic diagram of PEDF's effect on blocking Wnt3a, allowing for terminal differentiation of osteoblasts. In the absence of PEDF, unopposed Wnt3a exposure impedes terminal osteoblast differentiation and appropriate mineralization. PEDF potentiates osteoblast precursors to differentiate into mature osteoblasts. MSC, mesenchymal stem cell.

DISCUSSION

In the current study, PEDF restoration in a murine model of OI type VI increased trabecular and cortical bone mass and altered the mechanical properties of bones by making them more ductile. The findings have potential therapeutic implications for patients with OI type VI. Unlike OI due to collagen mutations (OI types I–IV), patients with OI type VI respond poorly to chronic bisphosphonate therapy. This is manifested by a higher fracture incidence and impaired mobility relative to patients with other OI subtypes, which reflects a persistent mineralization defect that was not altered by bisphosphonate therapy (17). Recent exome studies of previously uncharacterized autosomal recessive patients with OI indicate that a sizeable fraction harbor null mutations in PEDF (9, 32–34). Moreover, the discovery that a subtype of OI type V converges on nearly undetectable PEDF levels indicates a broader role for PEDF in bone development (10). This suggests that addressing PEDF deficiency may represent a potential modality for correcting defects in bone development and functionality in patients with OI type VI and possibly other OI subtypes.

PEDF restoration in null mice had trophic effects on bone mass in young and older mice, but differences were apparent. In young mice, an effect on trabecular number was prominent. In contrast, PEDF accentuated trabecular thickness in mature mice. These age-specific differences in anabolic effects may reflect normal bone development in young mice via mineralization of osteoid in the presence of PEDF. In older KO mice, abnormal bone formation occurred with abundant osteoid in the absence of PEDF, which, when PEDF was restored, mineralizes this osteoid, thereby creating thicker trabeculae. In addition, osteoclast activity may also be affected by PEDF. One study noted that PEDF increases osteoprotegerin, the RANK ligand decoy receptor, that mitigates osteoclast activity (35). Thus, future studies that investigate PEDF-directed effects on bone resorption by osteoclasts will further delineate the role of PEDF in bone remodeling.

Other investigators have also reported that PEDF protein directs hMSCs to the osteoblast lineage (15, 16). Li et al. (15) described PEDF-induced mineralization of bone marrow–derived MSCs and enhanced β-catenin levels. Our previous report that demonstrated PEDF stimulation of the Wnt coreceptor, LRP6, in early hMSCs was unexpected as PEDF inhibits Wnt signaling in eye cells (18). High levels of endogenous PEDF at the mid to late stages of osteoblast differentiation supported the possibility that PEDF could have distinct temporal effects on Wnt signaling (Fig. 4). To model late-stage differentiation effects, continuous Wnt3a exposure inhibited mineralization by nearly 50% in differentiating hMSCs and suppressed active β-catenin levels. Concurrent Wnt3a and a PEDF 34-mer also did not enhance hMSC mineralization; however, addition of PEDF 34-mer during the final 8 d of differentiation potentiated mineralization compared with Wnt3a-treated cells. Although Wnt ligands direct MSC lineage specification to the osteoblast lineage, unchecked Wnt3a impedes terminal differentiation to osteoblasts, which is consistent with the concept that pluripotency of select stem cell populations is Wnt dependent (28, 36). Presence of PEDF or PEDF peptide allows for terminal differentiation and mineralization, thereby providing a mechanism for the phenotype seen in OI type VI.

A binary differentiation niche in the bone marrow mediated by a gradient of canonical Wnt ligands that inhibits adipogenesis but is permissive for self-renewal and early differentiation to the osteoprogenitor state has been detailed (30). For progenitor cell differentiation to mature osteoblasts, exposure to Wnt ligands would dissipate and Wnt inhibitors would increase, allowing for terminal differentiation. Our results suggest a critical role for PEDF in this stem cell niche in which effects of Wnt ligands are balanced by multiple endogenous inhibitors of Wnt signaling. This is supported by the finding that PEDF can inhibit adipogenesis when promoting osteoblast differentiation (14, 37). Other Wnt antagonists, such as DKK-2, also function in an analogous capacity, enhancing early Wnt signaling but inhibiting the Wnt pathway during the final stage of osteoblast differentiation to increase mineralization (38, 39). The role of cellular context and the differential expression of receptors at various stages of differentiation may underlie Wnt agonist vs. antagonist activities for such ligands as PEDF (39, 40).

During submission of the current article, another group has published that PEDF restoration had no effect on bone mass in PEDF-null mice (41). Because our manuscript reaches the exact opposite conclusion, we address points that may explain these differences. Rajagopal et al. (41) used a viral vector delivery system that achieved serum PEDF levels >1000 µg/ml. These levels were reported in control vector–treated mice as well and are not within the normal physiologic range reported in mice or humans. PEDF levels were reported by Crowe et al. (42) on the order of ∼10 ng/ml in WT mice. We reported serum PEDF levels in the range of 75–120 ng/ml (19). In humans, PEDF levels have been reported to be ∼5–20 µg/ml (11, 43). Prior functional studies of PEDF biology found paradoxical effects with higher concentrations of PEDF (44). The age of mice in their studies was older (6 mo at death) and may have impacted their results. Thus, several factors differ from the present study, which found a trophic effect on bone mass by PEDF in PEDF-null mice.

In the adult organism, the cellular source of PEDF needed to maintain bone mass remains unclear. Treatment with systemic PEDF restoration was adequate to increase bone mass in vivo, which suggests that systemic circulating levels of PEDF are adequate to enhance bone mass. This is relevant as the normal liver is the major expression site of PEDF, with the recent Human Protein Atlas demonstrating PEDF expression by RNA sequencing of adult bone to be ∼1% of levels found in the liver (12, 45). Whether PEDF represents a liver-derived factor that contributes to bone development remains unclear and awaits experimentation by using liver-specific PEDF-KO mice.

Related to the source of PEDF is the common clinical finding that liver-related diseases in which PEDF is suppressed are associated with marked bone demineralization and fractures. For instance, chronic liver diseases, such as cholestatic disorders, alcoholic liver disease, and others, are characterized by a profound osteodystrophy. We previously showed that 2 models of chronic alcohol intake resulted in marked suppression of liver PEDF (23). We have found that ethanol feeding in PEDF-KO mice accentuates bone demineralization more than that seen in corresponding WT mice (unpublished data). Future studies that examine whether PEDF can rescue bone loss in such models as well as in those associated with aging and osteoporosis may provide further insight into its role in bone homeostasis in adult organisms.

In summary, PEDF reconstitution in a murine model of OI type VI had a trophic effect on bone mass and lessened the “brittle” in brittle bone disease. A conserved motif on the PEDF protein accentuates blockade of the Wnt receptor, LRP6, which indicates an inhibitory effect on Wnt signaling that allows for terminal osteoblast differentiation. These findings have potential therapeutic implications.

Glossary

µCT

micro–computed tomography

ATCC

American Type Culture Collection

BV/TV

bone volume/total volume

DKK

dickkopf

hMSC

human mesenchymal stem cell

HPMC

hydroxypropylmethylcellulose

KO

knockout

OI

osteogenesis imperfecta

PEDF

pigment epithelium–derived factor

WT

wild-type