Determination of mesenchymal stem cell fate by pigment epithelium-derived factor (PEDF) results in increased adiposity and reduced bone mineral content

Abstract

Pigment epithelium-derived factor (PEDF), the protein product of the SERPINF1 gene, has been linked to distinct diseases involving adipose or bone tissue, the metabolic syndrome, and osteogenesis imperfecta (OI) type VI. Since mesenchymal stem cell (MSC) differentiation into adipocytes vs. osteoblasts can be regulated by specific factors, PEDF-directed dependency of murine and human MSCs was assessed. PEDF inhibited adipogenesis and promoted osteoblast differentiation of murine MSCs, osteoblast precursors, and human MSCs. Blockade of adipogenesis by PEDF suppressed peroxisome proliferator-activated receptor-γ (PPARγ), adiponectin, and other adipocyte markers by nearly 90% compared with control-treated cells (P<0.001). Differentiation to osteoblasts by PEDF resulted in a common pathway that involved PPARγ suppression (P<0.01). Canonical Wnt-β-catenin signaling results in a MSC differentiation pattern analogous to that seen with PEDF. Thus, adding PEDF enhanced Wnt-β-catenin signal transduction in human MSCs, demonstrating a novel Wnt agonist function. In PEDF knockout (KO) mice, total body adiposity was increased by >50% compared with controls, illustrating its systemic role as a negative regulator of adipogenesis. Bones from KO mice demonstrated a reduction in mineral content recapitulating the OI type VI phenotype. These results demonstrate that the human diseases associated with PEDF reflect its ability to modulate MSC differentiation.—Gattu, A. K., Swenson, E. S., Iwakiri, Y., Samuel, V. T., Troiano, N., Berry, R., Church, C. D., Rodeheffer, M. S., Carpenter, T. O., Chung, C. Determination of mesenchymal stem cell fate by pigment epithelium-derived factor (PEDF) results in increased adiposity and reduced bone mineral content.

The differentiation of stem cells into distinct tissues can be regulated by specific extracellular signaling molecules. For example, Wnt ligands and VEGF differentially regulate mesenchymal stem cell (MSC) fate into adipocytes or osteoblasts (1–4). Two human diseases affecting adipocytes and osteoblasts, obesity and osteogenesis imperfecta (OI) type VI, have been associated with an excess or complete absence of pigment epithelium-derived factor (PEDF) (5–8). PEDF is a 50-kDa secreted multifunctional protein of the SERPIN superfamily that has been implicated in the regulation of stem cell populations (9–12). A prior study suggested that PEDF may induce osteoblast differentiation from embryonic stem cells, but PEDF dependency was not evaluated (12). Whether PEDF plays a direct role in the commitment and differentiation of MSCs into adipocytes or osteoblasts, the two cell types underlying the extremes of PEDF-related human diseases, has not been investigated.

The clinical manifestations of high PEDF vs. its absence point to its role in adipocyte and osteoblast development. Increased PEDF levels correlate with adiposity in patients with the metabolic syndrome (7, 8, 13). Here, elevated PEDF likely represents a compensatory measure since PEDF impedes adipogenesis of 3T3-L1 adipocyte precursors and its absence in mice results in ectopic lipid accumulation in organs such as the liver and pancreas (14–16). Conversely, individuals lacking PEDF because of null mutations have OI type VI, an autosomal recessive form of OI characterized clinically by fractures of bone due to inadequate mineralization (6, 17). Bone specimens from patients with OI type VI reveal severely hypomineralized bones that are mirrored in a mouse model of PEDF deficiency (18, 19). The mineralization defect was associated with abnormalities in the extracellular matrix that were reported in osteoblast cultures and bones from these mice (19). Although exome sequencing established null mutations in the PEDF gene as the cause of OI type VI, a mechanism for the phenotype remains unclear (5, 6, 17).

Our group previously reported obvious abnormalities of mesenchymal progenitor-derived cells in the livers and pancreas of PEDF knockout (KO) mice (16, 20). This included a striking pattern of α-smooth actin staining reflecting activation of mesenchymal progenitor-derived cells (16, 20). Also prominent was the marked presence of lipid droplet markers in PEDF KO fibroblasts in organs normally devoid of adipocytes (21). This finding suggested an adipogenic drive of mesenchymal cells in the absence of PEDF. Based on these previous findings, and that of the two clinical phenotypes associated with excess and absent PEDF, we investigated whether PEDF might affect these phenotypes by modulating MSC differentiation toward adipocytes vs. osteoblasts.

Using murine and human MSCs (hMSCs), we found that PEDF significantly inhibited adipogenesis while promoting osteogenesis. Since peroxisome proliferator-activated receptor γ (PPARγ) signaling can govern adipocyte vs. osteoblast differentiation and PEDF has been shown to regulate PPARγ (2, 14, 22–24), we investigated the ability of PEDF to modulate developmental pathways that control PPARγ expression. In fact, PEDF-directed MSC differentiation led to activated Wnt-β-catenin signaling and marked PPARγ suppression. The imbalance between adipogenesis and osteogenesis was reflected in PEDF KO mice, which showed enhanced adiposity and decreased bone mineral content, thereby capturing key phenotypic features of the metabolic syndrome and OI type VI.

MATERIALS AND METHODS

Chemicals

The inhibitor of Wnt production (IWP-2; Tocris Bioscience, Minneapolis, MN, USA; 2 μM × 24–48 h) was used to block production of endogenous Wnt proteins (25, 26). All other chemicals, unless indicated, were purchased from Sigma (St. Louis, MO, USA).

Animals

PEDF KO mice have been described previously (27). PEDF KO mice were bred with 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 a 12/12-h light-dark cycle and fed with a standard mouse chow diet ad libitum. Tibiae and femurs from 14- and 26-d-old mice were processed at the Yale Core Center for Musculoskeletal Disorders. A high-fat diet (45% calories from fat; Research Diets, New Brunswick, NJ, USA) was given for 1 wk where indicated. All animal experiments were done in accordance to the Yale University and Veterans Affairs Connecticut Institutional Animal Use and Care Committees.

Primary cells

Subcutaneous fat pads were dissected from WT and PEDF-KO mice and digested in HBSS medium containing 3% BSA (American Bioanalytical, Natick, MA, USA), 0.8 mg/ml of type 2 collagenase (Worthington Biochemical Corp., Lakewood, NJ, USA), 1.2 mM CaCl₂, and 1.0 mM MgCl₂ for 1 h and 15 min in a shaking 37°C water bath. Stromovascular fractions (SVFs) were obtained after centrifugation at 300 g for 5 min. Cells were initially plated in proliferation medium (DMEM plus 10% FBS and bFGF; 1 ng/ml) until 70–80% confluence. Adipogenic differentiation was initiated with differentiation medium (5 μg/ml insulin, 10 nM dexamethasone, 0.5 mM IBMX, and 1 μM rosiglitazone) and added on d 0 for 72 h. Afterward, cells were maintained in DMEM with 10% FBS and 5 μg/ml insulin for an additional 5 d until full differentiation as confirmed by light microscopy and Oil Red O staining. Osteoblast differentiation cocktail (10 nM dexamethasone, 50 μg/ml ascorbic acid, and 10 mM β-glycerophosphate) was used for 21 d.

Adipose-derived hMSCs [American Type Culture Collection (ATCC), Manasses, VA, USA; PCS-500-011) were propagated in MSC basal medium supplemented with MSC growth kit (ATCC). Cells were CD29, CD44, CD105, and CD166 positive and negative for CD 31, CD34, and CD45. Interrogation of Wnt-β-catenin signaling was done in ATCC MSC basal medium with 2% FBS. PEDF (500 ng/ml) was added for 6 h. Passage 2 and 3 hMSCs were used for Wnt signaling experiments. For adipocyte conversion, cells (5000 cells/cm²) were placed in adipocyte differentiation medium (ATCC; PCS-500-050) for 9–10 d. For osteoblast differentiation, cells (5000 cells/cm²) were differentiated in osteoblast differentiation medium (ATCC; PCS-500-052) for 10–21 d. Alkaline phosphatase kit (Sigma) was used to stain osteoblasts.

Osteoblast progenitor cells were isolated as described previously (28). Briefly, calvaria were dissected from WT and PEDF KO mice with careful removal of all visible connective tissue surrounding the calvaria. After being washed in HBSS, calvaria were digested in collagenase type 1 (1 mg/ml; Worthington Biochemical) in HBSS for 10 min in a shaking water bath at 37°C. The first collection of supernatant was discarded, and calvaria were digested for an additional 20 min with the second and third sets of supernatants collected. Cells were washed twice in 2% FBS. Osteoblast progenitors were differentiated into osteoblasts in osteoblast differentiation medium for 21 d unless specified. In vitro assays assessing adipocyte or osteoblast differentiation were performed ≥3 separate times, with n = 3–4 for each separate experiment.

Sorting of adipose populations

Excised white subcutaneous adipose tissue was digested in 0.8 mg/ml collagenase type 2 (Worthington Biochemical; LS004174) in HBSS containing 3% BSA, 1.2 mM CaCl₂, 1.0 mM MgCl₂, and 0.8 mM ZnCl₂ for 75 min in a shaking water bath. Floating adipocytes were separated from the SVF via centrifugation at 300 g for 3 min. Isolation of intact adipocytes was verified by staining with plasma membrane (Cell Mass Orange) and nuclear (DAPI) dyes. Purified adipocytes were then placed into TRIzol LS Reagent (Invitrogen, Grand Island, NY, USA; 10296028) for RNA isolation. SVF was sequentially filtered through 70- and 40-μm filters before staining with the following antibodies for 20 min: CD45 APC-eFluor 780 at 1:5000 (eBioscience, San Diego, CA, USA; 47-0451-80), CD31 PE-Cy7 at 1:1200 (eBioscience; 25-0311-82), CD29 Alexa Fluor 700 at 1:400 (BioLegend, San Diego, CA, USA; 102218), CD34 Alexa Fluor 647 at 1:200 (BioLegend; 119314), Sca-1 Pacific Blue at 1:1000 (BD Biosciences, San Jose, CA, USA; 560653). Following antibody incubation, samples were washed and centrifuged at 300 g for 3 min.

PEDF protein and PEDF restoration

Full-length recombinant human PEDF was generated in HEK cells (15). PEDF integrity was confirmed by silver staining (Invitrogen), and then dialyzed in PBS.

RNA analysis and quantitative reverse transcriptase polymerase chain reaction (qRT-PCR)

RNA was isolated and processed using the RNAEasy mini kit (Qiagen, Valencia, CA, USA). Primer probe sets (see Supplemental Data) were obtained from a commercial source (Applied Biosystems, Foster City, CA, USA), and qRT-PCR was performed on a TaqMan ABI 7500 system (Applied Biosystems) as described previously (20). Target gene expression was normalized against 18S ribosomal RNA using the ΔΔC_t method. On sorted adipose populations, qRT-PCR was performed on a Roche Lightcycler 480 (Roche, Basel, Switzerland) using a SYBR FAST qPCR kit (Kapa Biosystems, Woburn, MA, USA; KK4611), and target gene expression was normalized to TBK1 (see Supplemental Data).

Immunoblotting

Immunoblotting was performed as described previously (20). Proteins were separated by 10% SDS-PAGE on gradient gels (Bio-Rad, Hercules, CA, USA). Antibodies against nonphosphorylated β-catenin, total β-catenin, phospho-LRP6 (serine 1490), total LRP6 (Cell Signaling Technologies, Danvers, MA, USA), and alkaline phosphatase (Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, IA, USA). To assess signaling through the Wnt coreceptor LRP6, 3T3 adipocyte precursors with vector and LRP6-KO shRNA were used (gift of Dr Arya Mani, Yale University School of Medicine; ref. 29). On confluence, cells were serum starved for 48 h, and then changed to DMEM with 10% FCS (d 0). PEDF (10 nM) was added on d 0, and cells were harvested after 48 h. Knockdown was confirmed by immunoblotting for total LRP6. β-actin (Sigma) was used as a loading control.

Staining

For Oil Red O staining, cells were washed in PBS and fixed in 10% formalin for 20 min. Cells were then stained with 0.5% Oil Red O for 15 min. For Alizarin Red staining, cells were washed in PBS and fixed in 10% formalin for 20 min. After being washed twice in ddH₂O, cells were stained in 2% Alizarin Red S (pH 4.1). Cells were washed ≥3 times in ddH₂O, and images were acquired with a Zeiss Axiophot microscope (Carl Zeiss, Oberkochen, Germany). Representative images are shown at ×40 view.

Bone micro-computed tomography (microCT) and histological analysis

A ScanCo μCT 35 scanner (ScanCo Medical AG, Brüttisellen, Switzerland; Yale Center Core for Musculoskeletal Disorders) was used to assess the distal femur for trabecular and cortical bone morphology from 21-d-old WT and PEDF-KO mice in a blinded manner. Axial, sagittal, and coronal images were obtained at standardized sites, and measures of trabecular and total bone volumes and other parameters of bone density were obtained (30). For histology, femurs and tibiae of 14- and 26-d-old mice were dissected, cleaned, and fixed in 70% ethanol, then further dehydrated through graded ethanols, cleared in toluene, and embedded in methyl methacrylate (MMA). After polymerization, MMA 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 μm thickness were cut using a Reichert-Jung RM 2165 microtome (Leica Microsystems, Jena, Germany) using a D-profile tungsten carbide knife, mounted on charged slides, and stained with either Goldner's trichrome or Toluidine Blue O (pH 3.7).

Statistical analysis

Results were assessed 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 ± se. A value of P < 0.05 was considered significant.

RESULTS

PEDF inhibits adipogenesis by suppressing PPARγ and its coactivator peroxisome proliferator-activated receptor γ coactivator 1α (PGC1α)

We assessed the ability of PEDF to modulate murine and hMSC differentiation into adipocytes. MSCs derived from stromovascular cells (SVCs) from adult WT and PEDF KO mice were able to differentiate into adipocytes by d 8 (Fig. 1A). PEDF (500 ng/ml) added on d 1 or 3 significantly inhibited adipogenesis (Fig. 1B). This was accompanied by suppression of the proadipogenic transcription factors PPARγ, PGC1α, CCAAT/enhancer-binding protein-α (CEBPα), and the adipocyte marker ADIPOQ on d 1 (Fig. 1C) and d 3 (Fig. 1D). PPARγ suppression was greater when PEDF was started on d 1 compared with d 3. Even at this later time, PEDF inhibited PPARγ and PGC1α expression by nearly 50% in both WT and KO SVCs (Fig. 1D). PEDF similarly inhibited proadipogenic transcription factors in hMSCs (Fig. 1E). Thus, PEDF inhibits adipogenesis by inhibiting the key proadipogenic transcription factors, PPARγ and its coactivator PGC1α.

Figure 1.

PEDF suppresses adipogenesis. A) MSCs derived from SVCs from the subcutaneous fat pads of WT and PEDF KO mice undergo adipogenesis by d 8, as demonstrated by Oil Red O staining. B) MSCs from WT mice were treated with vehicle or PEDF (500 ng/ml/d) and assessed for adipocyte differentiation, represented by Oil Red O staining, on d 8. C) Treatment with PEDF starting on d 1 of differentiation significantly suppressed multiple proadipogenic transcription factors and adipocyte-specific markers. D) Proadipogenic transcription factor PPARg and its cofactor, PGC1a, were significantly higher in PEDF KO compared with WT MSCs at baseline. PEDF inhibited PPARg and PGC1a expression in WT and KO SVCs when PEDF was initiated on d 3 of differentiation. E) PEDF inhibits adipogenesis of hMSCs. PEDF was started on d 0 of differentiation, and cells were differentiated in adipogenic medium for 10 d. Ad, adipocyte differentiation medium; vehicle, PBS.

PEDF is suppressed during adipogenesis, and Wnt expression maintains PEDF expression

If PEDF were suppressed during adipogenesis, it would support its function as a negative regulator of adipocyte development. Data on PEDF expression by adipocyte precursors and adipocytes, however, are discordant. Some proteomic studies note high PEDF levels in preadipocytes that declined markedly early in adipogenesis, but others demonstrate high PEDF secretion by mature adipose tissue (31, 32). To clarify this issue, FACS-sorted adipocyte precursors (CD45⁻, CD31⁻, CD29⁺, CD34⁺, and Sca-1⁺) were compared with mature adipocytes for PEDF expression (33). Adipocyte progenitors had 8-fold higher PEDF levels compared with mature adipocytes (Fig. 2A). To evaluate our results in the context of previous analyses of adipogenesis, we analyzed PEDF expression during adipogenesis using unbiased microarray data (34). PEDF expression was similarly reduced by nearly 90% in SVCs undergoing adipogenesis (Fig. 2B), analogous to the results seen in our FACS-sorted study. Wnt agonists maintain preadipocytes in an undifferentiated state despite adipogenic stimuli (34). Under Wnt expression and adipogenic medium, PEDF levels were maintained to a greater extent than those cells without Wnt stimulation (Supplemental Fig. S1). FACS-sorted adipocyte precursors and unbiased microarray data demonstrate that PEDF expression is markedly suppressed to a similar extent after adipocyte differentiation. Wnt proteins, which inhibit adipogenesis, moreover, increase PEDF expression. These findings illustrate the ability of Wnt signaling to maintain antiadipogenic signals such as PEDF, which are suppressed during adipogenesis (14, 31).

Figure 2.

PEDF expression is markedly suppressed during adipogenesis. A) PEDF expression from FACS-sorted adipocyte precursors and mature adipocytes derived from subcutaneous fat. B) Analysis of unbiased microarray analyses of SVCs undergoing adipogenesis was interrogated for PEDF expression. PEDF expression was reduced by 90% after SVC conversion to mature adipocytes.

PEDF promotes mineralization and osteoblast differentiation in vitro

To define the role of PEDF in osteoblast mineralization and differentiation, SVCs from PEDF KO mice, hMSCs, and osteoblast progenitors were induced to undergo osteoblast differentiation in the absence and presence of PEDF. Exogenous PEDF treatment for 21 d increased Alizarin Red staining of KO SVCs above that found in vehicle-treated cells (Fig. 3A). PEDF did not affect the prototypical bone transcription factors, Runx2 (Fig. 3A) or Sp7 (data not shown), at d 21 of differentiation. Collagen Ia expression was also not different in PEDF-treated cells consistent with results noted in clinical OI type VI where the absence of PEDF was not associated with collagen mutations or processing (17). Comparable to the results seen in adipogenesis, PEDF markedly reduced PPARγ expression (Fig. 3A), a negative regulator of osteoblast differentiation, by nearly 80% (2). PEDF also reduced the expression of other negative regulators of bone formation, such as transforming growth factor β (TGF-β; Fig. 3A and refs. 35, 36). Thus, PEDF promotes osteoblast differentiation by suppressing negative regulators of bone formation.

Figure 3.

PEDF enhances osteoblast mineralization and differentiation. A) PEDF KO SVCs were placed in osteoblast differentiation medium for 21 d and stained with Alizarin Red (top panels). KO SVCs were treated with vehicle or PEDF (500 ng/ml/d) starting on d 2 of osteoblast differentiation, and gene expression evaluated on d 21. Differences in Runx2 or collagen 1a1 expression were not seen with PEDF, while PEDF significantly suppressed TGF-β and PPARγ expression. B) hMSCs demonstrate increased alkaline phosphatase staining with PEDF. C) Osteoblast progenitors from WT and PEDF KO mice display increased alkaline phosphatase staining in response to PEDF. Os, osteoblast differentiation medium.

We next determined whether PEDF could differentiate hMSCs and osteoblast progenitors toward the osteoblast lineage. Adding PEDF to hMSCs increased alkaline phosphatase, a marker of osteoblast differentiation, by staining and protein levels (Fig. 3B). In committed osteoblast progenitors, PEDF KO cells demonstrated reduced alkaline phosphatase intensity compared with WT cells at baseline (Fig. 3C). Gene expression confirmed diminished alkaline phosphatase expression in PEDF KO compared with WT cells without differences in collagen 1A1 (Supplemental Fig. S2). In addition, thrombospondin 1 (TSP-1), a negative regulator of late osteoblast maturation (35), was increased in KO cells (Supplemental Fig. S2). Restoring PEDF (50 ng/ml) to KO osteoblast progenitors resulted in alkaline phosphatase intensity that was similar to untreated WT cells (Fig. 3C). With higher PEDF concentrations (500 ng/ml), alkaline phosphatase staining was similar between WT and PEDF KO cells (Fig. 3C). These results indicate that PEDF can induce osteoblast differentiation in SVCs, hMSCs, and committed osteoblast progenitors. Moreover, PEDF-mediated blockade of adipogenesis and promotion of osteogenesis were associated with marked PPARγ suppression, indicating a common pathway by which PEDF exerts differentiation of MSCs.

PEDF acts as a Wnt agonist for human MSCs

Wnt-β-catenin signaling can determine osteoblast vs. adipocyte specification through suppression of transcription factors PPARγ and CEBPα (1, 2, 37). Since PEDF treatment resulted in a transcriptional profile akin to Wnt activation (Figs. 1C and 3A), we interrogated the ability of PEDF to modulate canonical Wnt-β-catenin signaling including the activation (phosphorylation) status of the Wnt cell surface receptor LRP6 and the ratio of active (nonphosphorylated) β-catenin to total β-catenin. The prototypical agonist, Wnt3a, resulted in increased LRP6 phosphorylation of hMSCs (Fig. 4A). Similarly, PEDF treatment of hMSCs led to increased LRP6 phosphorylation (Fig. 4A). Since PEDF had functional effects that indicated it acts as a Wnt agonist but only minimal effects on active β-catenin levels under medium conditions where endogenous Wnt proteins would be present (data not shown), we assessed whether PEDF functions directly on LRP6 activation or acts indirectly by increasing endogenous Wnt production. To remove endogenous Wnt production, hMSCs were preincubated with the potent small molecule IWP-2 for 24–48 h (25). IWP-2 incubation (48 h) alone effectively blocked endogenous Wnt production as evidenced by near absence of LRP6 phosphorylation (Fig. 4B). The addition of PEDF to IWP-2-treated cells led to LRP6 phosphorylation and enhanced levels of active β-catenin (Fig. 4B). Thus, PEDF functions as a direct Wnt agonist in hMSCs.

Figure 4.

A) PEDF activates Wnt signaling in hMSCs. hMSCs were treated with Wnt3a (50 ng/ml) and PEDF (500 ng/ml) and immunoblotted for phosphorylated LRP6. B) hMSCs were treated with IWP-2 (2 μM) for 24 and 48 h and then challenged with PEDF (500 ng/ml in basal medium). Blots are representative of n = 4 experiments/condition. Vehicle was PBS for Wnt3a and PEDF experiments; DMSO for IWP-2 experiments. C) Committed preadipocytes, 3T3-L1 cells, transfected with vector and shRNA targeting LRP6 were assessed for knockdown of LRP6. D) Adding PEDF significantly suppressed PPARγ in vector-transfected cells, while LRP6 knockdown resulted in increased PPARγ expression.

We next evaluated whether the suppressive effect of PEDF on PPARγ expression was LRP6 dependent. Knockdown of total LRP6 was confirmed in 3T3-L1 adipocyte precursors (Fig. 4C). In control (vector) cells, PEDF significantly inhibited PPARγ expression despite commitment of these cells to the adipocyte lineage (Fig. 4D). In contrast, PEDF induced PPARγ expression in cells with LRP6 knockdown (Fig. 4D), demonstrating that the suppressive actions of PEDF on PPARγ expression is LRP6 dependent.

PEDF deficiency in mice results in increased total body fat and altered bone mineralization

Given human studies showing elevated circulating PEDF in the setting of obesity, it may have been predicted that PEDF KO mice would be lean. However, 12-wk-old PEDF KO mice exhibited a nearly 50% increase in total body adiposity as determined by MR spectroscopy but similar body weights (Fig. 5A). Dissection of subcutaneous, epididymal, and retroperitoneal fat pads in 12-wk-old mice confirmed increased adiposity with PEDF deficiency (Fig. 5B). Differences in fat depot size occurred under both normal and high fat diets (Fig. 5B). These findings indicate that absence of PEDF is permissive for increased adipogenesis in vivo, illustrating its systemic role as a negative regulator of adipogenesis.

Figure 5.

PEDF deletion is associated with increased total body adiposity and reduced bone mineral content in mice. A) Body weight and percentage of total body fat by MR spectroscopy of 12-wk-old WT and PEDF KO mice. B) Representative images of WT and PEDF KO subcutaneous (top left panel) and epididymal (bottom left panel) fat pads, and corresponding quantification of subcutaneous white adipose tissue (SWAT), epididymal white adipose tissue (EWAT), and retroperitoneal white adipose tissue (RWAT) under normal feeding and 1 wk of a high-fat diet (right panels). C) MicroCT-obtained images of trabecular, dorsal, and lateral surface bone morphology of distal femurs: left panels, cross-section; left center panels, dorsal frontal view; right center panels, left lateral surface; right panels, cut left lateral view. D) Quantification of trabecular bone volume (BV), total volume (TV), and BV/TV, demonstrating diminished trabecular volumes in PEDF KO bones. E) Low-power (4×) and high-power (10×) images of Goldner's stained tibiae and femurs from WT and PEDF KO mice. Decreased epiphyseal and chondro-osseous mineral content in 14-d-old KO compared with WT mice (arrows). Older (26-d-old) mice demonstrate hypomineralization in the epiphysis (short arrows) and chondro-osseous junction (longer arrows) with a diminished proliferative (P) zone; n = 6–9 mice for adipose tissue determination and n = 3–4 mice for bone imaging.

To evaluate whether PEDF deficiency in mice recapitulates the PEDF null mutation seen in patients, bone volumes and mineralization were examined in femurs and tibiae at multiple time points. MicroCT analysis showed decreased trabecular bone volume along the distal femur (Fig. 5C; left panels, cross section of trabecular volume; left center panels, dorsal frontal view). Surface and cut images along the distal femur further illustrated the decrease in bone density in PEDF KO mice (Fig. 5C; right center panels, left lateral surface; right panels, cut left lateral view). Trabecular bone volume (BV) in PEDF KO mice (Fig. 5D) was >40% less than in control mice, while the total volume (TV) between groups was not different. Thus, the diminished bone volume fraction (BV/TV) in PEDF KO mice primarily reflected the loss of trabecular volume. The deficiency in the trabecular bone volume of KO mice was further evident in a diminished connectivity density (Supplemental Fig. S3). Goldner's staining revealed decreased mineral content in the epiphysis and the chondro-osseous junction of 14-d-old mice (Fig. 5E, arrows). At 26 d, mineralization was evident in the epiphysis of PEDF KO mice, but the growth plate was smaller and less organized in PEDF KO mice compared with controls (Fig. 5E). Specifically, the zone of proliferating chondrocytes was diminished in PEDF KO mice (Fig. 5E). Thus, PEDF appears necessary for normal bone formation in mice, with PEDF deficiency recapitulating the defective bone volumes and hypomineralization seen in human OI type VI.

DISCUSSION

Human diseases with distinct phenotypes reflecting the extremes of PEDF expression demonstrate the new role we have described for this factor in regulating MSC differentiation fate to adipocytes or osteoblasts. PEDF markedly inhibited the adipogenic drive of SVCs and hMSCs when given at the early phases of adipocyte differentiation. This effect coincided with suppression of the prototypical adipogenic transcription factors PPARγ and its coactivator PGC1α and is analogous to an antiadipocyte differentiation effect that was also confined to an early time point in 3T3-L1 preadipocytes (14). In contrast, PEDF promoted osteoblast differentiation of MSCs and committed osteoblast progenitors. The functional effects of PEDF were accompanied by suppression of PPARγ and other inhibitors of osteoblast differentiation such as TGF-β and TSP-1 (2, 35). Furthermore, we demonstrated that PEDF functions as a Wnt agonist in hMSCs and that its suppressive action on PPARγ is LRP6 dependent, thereby linking canonical Wnt-β-catenin signaling with the ability of PEDF to suppress PPARγ.

Previous studies of the PEDF KO phenotype provided insights into a potential role involving MSC differentiation. The first study described abnormalities in the matricellular compartment with stromal expansion in the prostate and pancreas (27). In subsequent studies, we showed activation of mesenchymal progenitor-derived cells in the pancreas and livers in the absence of injury, with accentuated fibrotic responses on injury (16, 20). Robust staining for TIP47, a lipid droplet marker, and the presence of increased stromal adiposity indicated an adipogenic drive in the mesenchymal cell population of organs that are typically devoid of adipocyte infiltration (21). In the current study, we describe a bone defect that captures key aspects of human OI type VI and increased total body adiposity in the absence of PEDF implicates its role in directing MSC fate toward osteoblasts and away from adipocytes (Fig. 6).

Figure 6.

PEDF directs MSC fate toward osteoblasts and away from adipocytes. This occurs through its action as a Wnt-β-catenin agonist that suppresses PPARγ.

Stimulation of LRP6 phosphorylation by PEDF is consistent with the known effects of Wnt signaling on promoting osteogenesis at the expense of adipogenesis (1, 37). However, the ability of PEDF to act as an LRP6 agonist is in contrast to a previous study by Park et al. (38) that demonstrated that PEDF functions as a Wnt antagonist. However, their results were noted in nonpluripotent cells of the eye, where comprehensive studies showed PEDF avidly binds LRP6 and prevents β-catenin nuclear translocation (38). Previous microarray analysis examining the effects of Wnt expression on undifferentiated cells lends evidence to the notion that PEDF is involved in Wnt signaling (34). Wnt expression and PEDF both impede adipogenesis that is temporally restricted to the undifferentiated state (14, 34), with PEDF expression being Wnt dependent (Supplemental Fig. S1). This suggests a positive feedback loop to prevent adipocyte differentiation. Thus, the developmental status of the target cell appears to be a critical factor in determining whether PEDF acts as a Wnt agonist vs. antagonist.

The current results provide further evidence to implicate the role of PEDF in stem cell biology (9–11). A proteomic screen of 806 secreted proteins found that PEDF was able to preserve hESC pluripotency without factors such as bFGF; implantation of PEDF-maintained hESCs developed into teratomas in vivo (11). Knockdown of another proposed PEDF receptor, adipose triglyceride lipase (ATGL), in hESCs triggered the loss of Oct4 expression and led to cellular differentiation (11). In the murine brain, PEDF secreted by cells of the subventricular zone were able to maintain neuronal stem cell renewal through activation of Notch signaling (9). Moreover, in this and a prior study, we showed that PEDF can negatively regulate TSP-1 levels (20). A recent study detailed that TSP-1 signaling suppresses c-Myc expression, thereby promoting cellular differentiation (39). The ability of PEDF to negatively regulate TSP-1 may therefore represent another mechanism by which PEDF modulates stem cell populations.

The reciprocal regulation of adipocyte vs. osteoblast differentiation by PEDF highlights the role of matricellular proteins in metabolic and bone homeostasis. For example, other matrix proteins such as secreted protein acidic and rich in cysteine (SPARC) can impede adipogenesis through activation of Wnt-β-catenin signaling, with SPARC-deficient animals displaying increased adipose mass and osteopenia (40, 41). Similarly, PEDF contains binding sites for extracellular matrix constituents including collagen and heparin sulfate that likely play a role in its regulation and function (42). The ability of PEDF to modulate Wnt-β-catenin signaling and suppress TSP-1 and TGF-β indicates a multifaceted regulation of MSC differentiation that likely involves regulation of other matricellular proteins.

Determining the degree of adiposity in those OI patients with the PEDF-null mutation would provide additional data to support the findings in the current study. A recent study comparing patients with OI type VI to patients with other OI subtypes (I, III, and IV) and healthy control subjects indicated that PEDF deficiency was significantly associated with increased body mass index (BMI) (43). The increased BMI in patients with OI type VI, the results of the current study, and the decreased body weights seen in PEDF-overexpressing mice support the notion that PEDF regulates body mass (38, 43). These findings also suggest that elevated PEDF levels in the metabolic syndrome represent a homeostatic mechanism to coordinately modulate adipogenesis and ensure adequate bone mass (13, 44). Studies examining the metabolic phenotype of patients with OI type VI will likely shed new information on the role of in adipogenesis and metabolism.

The findings presented in this study have additional clinical implications. For instance, the normal liver is a major source of circulating PEDF (45, 46), and chronic liver diseases such as alcoholic steatosis and cirrhosis and primary biliary cirrhosis are characterized by a profound osteodystrophy. Our group previously published that chronic ethanol ingestion in two rodent models significantly depletes hepatic PEDF levels (16). Whether liver disease models that deplete hepatic PEDF lead to impaired osteoblast differentiation and a hepatic osteodystrophy-like phenotype is an area of investigation.

In summary, two human disease phenotypes associated with elevated or absent PEDF, the metabolic syndrome (increased adiposity) and OI type VI, are reflected in its ability to modulate MSC fate. The ability of PEDF to promote osteoblast differentiation suggests that elevated PEDF in the metabolic syndrome can be viewed as a regulatory mechanism to promote osteoblast differentiation in the setting of increasing body mass.