High Levels of Pigment Epithelium–Derived Factor in Diabetes Impair Wound Healing Through Suppression of Wnt Signaling

Diabetes, which is associated with multiple vascular complications, affects ∼170 million patients worldwide (1). Diabetic foot ulcer (DFU), which affects 15% of diabetic patients, is a common vascular complication that might lead to amputation (2).

The pathological impairments of wound healing are the main reason for amputation. The dynamic process of wound healing has been divided into the following three phases: inflammation, re-epithelization and neovascularization, and remodeling (3,4). Neovascularization, including angiogenesis and vasculogenesis, plays a significant role in skin maintenance and repair, relying on the participation and coordination of various cells, especially endothelial progenitor cells (EPCs) and endothelial cells (ECs) (5).

EPCs are a subpopulation of bone marrow cells, which are mobilized into peripheral blood and are recruited to ischemic sites, where they differentiate in situ into mature ECs, participating in the process of neovascularization (6,7). A previous study (8,9) identified that specific cell surface markers of EPCs, such as CD34, CD133, and Flk-1, are present in the cells participating in vasculogenesis. The mobilization of EPCs is initiated by the key regulator vascular endothelial growth factor (VEGF), which is secreted by multiple cell types, including macrophages, fibroblasts, and epithelial cells, during wound-induced hypoxia (1). VEGF induces the activation of endothelial nitric oxide (NO) synthase, subsequently leads to the production of NO (10,11), and administers stem cell factor (SCF) binding to c-kit–positive progenitor cells (12,13), orchestrating the mobilization of EPCs into the bloodstream. Furthermore, the recruitment of EPCs to the wound site depends on ischemia-induced upregulation of stromal cell–derived factor-1 (SDF-1), which is secreted by epithelial cells and myofibroblasts (1). Previous studies (14,15) have shown that the downregulation of VEGF results in impaired mobilization and functions of EPCs. In addition, the decreased expression of SDF-1 has also been suggested to contribute to deficient recruitment of EPCs (14).

Pigment epithelium–derived factor (PEDF) is a 50-kDa secreted glycoprotein in the serine proteinase inhibitor (Serpins) family (16). PEDF has a broad spectrum of activities, including neuroprotection and regulation of lipid metabolism, and is upregulated in insulin-resistant and obese human subjects (17,18). In addition, PEDF is known to have a potent antiangiogenic activity (19), counterbalancing the proangiogenic activity of VEGF (20). Our previous studies (21,22) have shown that PEDF levels are increased in the circulation of type 2 diabetic patients and in type 1 diabetic patients with microvascular complications. These observations suggest that the elevated circulating levels of PEDF in patients with diabetes are probably compensatory responses to diabetic retinopathy and nephropathy. Although the effect of PEDF on corneal epithelial wound healing has been reported (23), the role of PEDF in diabetic skin wound healing has not been documented.

Our recent study (24) showed that PEDF functions as an endogenous inhibitor of the Wnt pathway. Wnts are a family of evolutionarily conserved, secreted glycoproteins modulating wound healing (24,25). In the canonical Wnt signaling pathway, Wnt ligands bind with their receptors Frizzled and the coreceptor LDL receptor–related protein 6 (LRP6), resulting in phosphorylation of LRP6, and preventing phosphorylation and degradation of the transcriptional factor β-catenin. The stabilized β-catenin translocates into the nucleus, and associates with T-cell factor to activate Wnt target genes (26). Wnt signaling plays a pivotal role in numerous processes, including inflammatory responses (27), carcinogenesis, fibrosis, and angiogenesis (28–32). Furthermore, our previous study (33) first verified that the Wnt signaling pathway plays the regulatory role in the release of EPCs during retinal neovascularization in the oxygen-induced retinopathy model. Nevertheless, the role of PEDF in regulating EPCs through the Wnt signaling pathway has not been studied. Kallistatin, another member of the serpin family, with a potent antiangiogenic effect, impairs skin function and repair through suppression of the Wnt pathway (34). However, the modulation of PEDF in Wnt signaling during wound healing has not been verified.

In the current study, we investigate the role of PEDF in the regulation of wound healing through the Wnt signaling pathway. We also tested the hypothesis that PEDF impairs the numbers and functions of circulating EPCs by suppressing Wnt signaling, consequently delaying the process of wound healing.

The collection of the human samples was in strict agreement with the institutionally approved guidelines by the Second Affiliated Hospital of Sun Yat-Sen University, and each participant gave written informed consent. Type 2 diabetes was clinically diagnosed and confirmed in all patients with and without DFUs. Blood was collected from 30 diabetic patients with DFUs, 30 diabetic patients without DFUs, and 25 nondiabetic subjects. Clinical data of the subjects are presented in Supplementary Table 1.

PEDF levels in human plasma were quantified by ELISA (Millipore, Billerica, MA). The factors of mice plasma were determined using commercial ELISA kits for PEDF (Cusabio Biotech China), VEGF (Bio Excel, Guangzhou, People’s Republic of China), SCF (RayBiotech, Norcross, GA), total NO, and Nitrate/Nitrite (R&D Systems, Minneapolis, MN).

All of the animal experiment procedures were approved by the Institutional Animal Care and Use Committee of the Sun Yat-Sen University and University of Oklahoma Health Sciences Center. The strain of mouse used for the type 1 diabetes model, the db/db mouse, the PEDF knockout (KO) mouse, and the Wnt reporter mouse were all B6. For the diabetic mouse model, mice (6 weeks old) were fed a high-fat diet (HFD) for 4 weeks (diet D12492 [60% calories from fat]; Animal Center of Guangdong Province, Guangzhou, People’s Republic of China). Then, the mice received daily intraperitoneal injections of streptozotocin (STZ; 40 mg/kg/day) for 7 days to induce diabetes. The db/db mice were purchased from the Model Animal Research Center of Nanjing University (Nanjing, People’s Republic of China). The PEDF KO mice were provided as a gift by Dr. S.J. Wiegand (Regeneron Pharmaceuticals, Inc., Tarrytown, NY) (35), and Wnt reporter mice were obtained from The Jackson Laboratory and were bred in the Rodent Barrier Facility of the University of Oklahoma Health Sciences Center. Recombinant glutathione S-transferase (GST)-PEDF (PEDF) and a PEDF-neutralizing antibody (PEDF Ab) were obtained as previously described (18). Mice were injected intraperitoneally with recombinant PEDF (2 mg/kg) daily for 7 days, and the control group was injected with the same concentration of GST. For neutralizing PEDF in diabetic mice, mice were injected intraperitoneally with the PEDF Ab (GenScript China) 0.4 mg/kg/day, for 15 days, and the control group was injected with the same concentration of nonspecific IgG (Beyotime Institute of Biotechnology, Shanghai, People’s Republic of China). Adenovirus expressing PEDF (Ad-PEDF) was generated from the full-length human PEDF cDNA as previously described (35), and adenovirus expressing green fluorescent protein (Ad-GFP) was used as the control virus. The viruses were injected into the wound bed (5 × 10⁷ plaque-forming units/wound) at day 1 after the wound was generated. Lithium chloride (LiCl; Sigma-Aldrich, St. Louis, MO) was injected intraperitoneally (100 mg/kg) into 3-month-old diabetic mice daily for 7 days, with the same concentration of NaCl as in the control, as described previously (33). The metabolic phenotypes of PEDF KO mice and mice with type 2 diabetes with PEDF Ab are presented in Supplementary Table 2.

Dorsum was clipped in anesthetized mice, and then standardized circular wounds were made with a full-thickness 6-mm skin punch (Acuderm, Fort Lauderdale, FL). Wound closure rates were measured by tracing the wound area every day through taking photos and were quantified by using ImageJ software (National Institutes of Health). The frozen slides of wound tissue were stained with CD31 antibody (BD Biosciences, Franklin Lakes, NJ). The slides were examined under ×100 magnification to identify the area with the highest vascular density, and five randomized high-power field areas of the highest microvessel density (MVD) were selected for counting under ×200 magnification. The average of the five areas was recorded as the MVD level of this case.

Anesthetized mice were fixed in a modified centrifuge tube using mildly adhesive tape, with mice kept at a temperature of ∼32°C, and the laser Doppler probe was fixed firmly to skin to measure perfusion units using the Perimed PeriFlux System 5000 (Perimed, Stockholm, Sweden).

Bone marrow cells were isolated from the femurs and tibias of individual mice. Briefly, the bones were carefully crushed in M199 medium (Invitrogen, Gaithersburg, MD) after removal of the muscle around the bone, followed by removal of the red blood cells and washing. Then the cells were grown in a fibronectin-coated dish in the presence of 10% FBS (Invitrogen) in EGM-2 medium (containing VEGF, hydrocortisone, human fibroblast growth factor, IGF, ascorbic acid, human epidermal growth factor, and heparin; Lonza) (36).

EPCs were cultured in EGM-2 medium, and the culture medium was replaced at day 4 in culture to remove nonadherent cells. EPCs were used for function assays at day 7. For the migration assay, 10% FBS in 600 μL EGM-2 medium was placed in the lower chamber of 12-well plates, and then the cells (1 × 10⁴ cells) were plated on the upper chamber with 200 μL EGM-2 medium including 1% FBS for 12 h. The cells were fixed with 4% paraformaldehyde and stained with crystal violet (Sigma-Aldrich). For the adhesion assay, 96-well plates were coated with fibronectin (Sigma-Aldrich) for 45 min at 37°C with 5% CO₂. Then the plates were washed three times with PBS to remove the nonadherent cells. The adherent cells were fixed with 4% paraformaldehyde and stained with crystal violet. For tube formation assay, 48-well plates were coated with Matrigel (BD Biosciences). Bone marrow–derived EPCs (2 × 10⁴ EPCs/well) were plated in 150 μL EGM-2 medium and incubated at 37°C with 5% CO₂ for 12 h.

To quantify circulation and bone marrow EPCs by FACS analysis, cells were acquired from either 100 μL of peripheral blood or bone marrow. The cells were incubated for 1 h on ice, and were stained with a fluorescein isothiocyanate–conjugated anti-mouse CD34 antibody, phycoerythrin-conjugated anti-mouse CD133 antibody, and PerCP(cy5.5)-conjugated anti-mouse Flk-1 antibody. All of the antibodies were purchased from BD Biosciences.

Wound skin tissue samples obtained from within a circle ∼5 mm in diameter around the wound were isolated from six mice and were lysed with 500 μL 1× SDS buffer for the total protein extraction. Western blot analysis was performed as described previously (37,38). Antibodies for phosphorylated (p) LRP6 and non-p-β-catenin were purchased from Cell Signaling Technology (Danvers, MA). Antibodies for VEGF and SDF-1 were purchased from Santa Cruz Biotechnology (Dallas, TX). The same blot was stripped and reblotted for an antibody for β-actin (Sigma-Aldrich).

Skin and wounds from BAT-gal transgenic mice were fixed for 2 h in 4% paraformaldehyde, stained with 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-gal; Sigma-Aldrich), and incubated at 37°C for 12 h according to the manufacturer’s instructions (Sigma-Aldrich).

For all studies comparing more than two groups, one-way ANOVA using SPSS version 13.0 software was used, followed by least significant difference t test for multiple comparisons among groups. The Student t test was used for statistical analysis between two groups. Statistical significance was considered at P < 0.05. All data were expressed as the mean ± SD.

As shown in Fig. 1A, circulating levels of PEDF were significantly higher in diabetic patients with DFU (DM [diabetic patients without DFU] + DFU; 7.0 ± 0.6 µg/mL, P < 0.01) compared with diabetic patients without DFU (DM; 6.1 ± 0.7 µg/mL) and nondiabetic control subjects (NDM; 4.0 ± 0.5 µg/mL). Circulating EPCs, identified by CD34 and Flk-1 costaining, were counted by FACS. The numbers of circulating CD34⁺/Flk-1⁺ cells were significantly lower in the DM+DFU group (0.03 ± 0.02%, P < 0.01) compared with the DM (0.05 ± 0.01%) and NDM (0.09 ± 0.02%) groups (Fig. 1B). Furthermore, correlation analysis showed that circulating PEDF levels were negatively associated with numbers of CD34⁺/Flk-1⁺ cells (Fig. 1C; r = −0.76, P < 0.01) in DFU patients (Fig. 1C). HbA_1c levels were significantly higher in diabetic patients than nondiabetic individuals, but there was no difference between the DM+DFU and DM groups (Fig. 1D). Collectively, circulating PEDF levels were elevated in diabetic patients with DFUs, which were associated with the decreased numbers of circulating CD34⁺/Flk-1⁺ cells.

Plasma PEDF levels were significantly higher in 3-month-old db/db mice (7.85 ± 2.41 ng/mL, P < 0.01) compared with their NDM littermates (3.31 ± 1.11 ng/mL) (Fig. 2A). HFD- and STZ-induced DM mice also showed increased PEDF levels compared with age-matched NDM mice (P < 0.01) (Supplementary Fig. 1A). Plasma PEDF levels were upregulated in diabetic mice, which is consistent with the clinical data from diabetic humans.

The PEDF Ab accelerated wound closure in db/db mice (Fig. 2B and C) and DM mice (Supplementary Fig. 1B and C) compared with IgG. To further verify the role of PEDF in wound healing, diabetes was induced in PEDF KO and wild-type (WT) littermates by HFD in combination with STZ injection. The wound-healing delay was ameliorated in PEDF KO (DM) mice compared with WT (DM) mice (Fig. 2D). Taken together, our results showed that wound healing in diabetic mice was accelerated via blocking PEDF.

Wound closure in NDM mice treated with recombinant PEDF lagged behind the control group with protein GST (Fig. 2E). Moreover, we measured the rate of wound closure in nondiabetic PEDF KO and WT mice. The result showed that the wound healing of PEDF KO mice was faster compared with WT littermates (Fig. 2F). Our observations suggested that PEDF administration alone impairs wound healing in nondiabetic mice.

Skin MVD was measured, as PEDF plays a significant role as an angiogenesis inhibitor (39). Compared with the IgG control group, the PEDF Ab significantly increased microvascular density in the skin of db/db mice (Fig. 3A and B). In contrast, recombinant PEDF administration significantly reduced vascular density in NDM mice, compared with GST-treated controls (Fig. 3B). In NDM or DM conditions, PEDF KO mice had a higher vascular density in the skin compared with WT littermates (Fig. 3B). Further, PEDF KO mice developed significantly increased dorsal skin thickness (Fig. 3C and D) and higher skin blood flow rate (Fig. 3E and F) compared with WT mice. Moreover, PEDF Ab increased dorsal skin thickness in db/db mice compared with IgG (Supplementary Fig. 2A). Taken together, our results showed that PEDF inhibited angiogenesis, decreased the dorsal skin thickness, and impeded skin blood flow.

Although PEDF is a well-known antiangiogenic factor, its effect on EPC mobilization has not been documented. The PEDF Ab significantly increased the numbers of CD34⁺/Flk-1⁺ and CD133⁺/Flk-1⁺ cells in the peripheral blood of db/db mice, compared with those in the IgG-treated group (Fig. 4A–E). In contrast, recombinant PEDF decreased the numbers of CD34⁺/Flk-1⁺ and CD133⁺/Flk-1⁺ cells in NDM mice compared with those in the GST-treated control (Fig. 4F). PEDF KO mice had more CD34⁺/Flk-1⁺ and CD133⁺/Flk-1⁺ cells in the circulation than the WT littermates (Fig. 4G and H). However, the injection of PEDF and the PEDF Ab had no effect on the numbers of CD34⁺/Flk-1⁺ and CD133⁺/Flk-1⁺ cells in the bone marrow (Supplementary Fig. 3A and B). Our observation suggested that PEDF decreased the number of CD34⁺/Flk-1⁺ and CD133⁺/Flk-1⁺ cells in the peripheral blood of mice.

To determine the effect of the PEDF Ab on the function of EPCs, we measured the migration, adhesion, and tube formation of bone marrow–derived EPCs ex vivo (Fig. 4I, M, and Q). In a transwell migration assay, the cells isolated from the db/db mice with the PEDF Ab treatment showed substantially higher migration than those from IgG-injected mice (Fig. 4J). On the other hand, recombinant PEDF injection slowed the migration of isolated EPCs, compared with GST injection in NDM mice (Fig. 4K). The EPCs from PEDF KO mice showed a higher migration activity than EPCs from WT mice (Fig. 4L).

In an adhesion assay, the PEDF Ab improved the adhesion function of the EPCs from db/db mice (Fig. 4N). Recombinant PEDF reduced the adhesion of EPCs, compared with the GST control in NDM mice (Fig. 4O). The EPCs from PEDF KO mice had more cells adhere to the plate compared with WT control mice (Fig. 4P).

In a tube formation assay, the PEDF Ab increased the tube and branch formation of EPCs from db/db mice (Fig. 4R). Recombinant PEDF reduced the tube formation of EPCs, compared with the GST-injected control mice (Fig. 4S). The EPCs from PEDF KO mice formed more tubes and branches than the WT mice (Fig. 4T). Moreover, PEDF Ab increased circulating levels of EPCs and improved the functions of EPCs in DM mice compared with IgG (Supplementary Fig. 4A–D). Our results showed that the function of EPCs was improved by neutralizing PEDF.

The expression of VEGF and SDF-1 was significantly higher in the protein levels seen in db/db mice treated with the PEDF Ab, compared with IgG, in the wound tissue at day 15 after wounding (Fig. 5A–C). In contrast, PEDF downregulated VEGF and SDF-1 expression in the wound tissue at day 7 after wounding (Fig. 5D–F). PEDF KO mice had higher VEGF and SDF-1 expression in the wound tissue than WT littermates (Fig. 5G–I). VEGF levels were significantly higher in db/db mice treated with the PEDF Ab (152.8 ± 17.5 pg/mL) compared with diabetic mice treated with IgG (102.1 ± 17.3 pg/mL; Fig. 5J). Plasma VEGF levels were significantly lower in NDM mice that had received the PEDF treatment (82.0 ± 11.7 pg/mL) compared with NDM mice that had received GST treatment (138.1 ± 15. 3 pg/mL; Fig. 5J). Similar to the change in VEGF levels, plasma SCF and NO levels were increased in db/db mice after treatment with the PEDF Ab, and decreased after the injection of recombinant PEDF (Fig. 5K and L). Moreover, PEDF Ab treatment increased the plasma levels of VEGF, NO, and SCF (Supplementary Fig. 6A–C), and VEGF and SDF-1 expression in wound tissue (Supplementary Fig. 6D) in DM mice compared with IgG treatment. We also measured the content of NO, which functions to induce EC migration and proliferation, in the wound tissue (40). The PEDF Ab increased the local NO level compared with diabetic mice treated with control IgG (Supplementary Fig. 7A and B). NO content in the wound tissue was decreased by the injection of recombinant PEDF compared with the NO content in the GST-injected control group (Supplementary Fig. 7C). Our observation suggested that PEDF decreases the expression of VEGF and SDF-1 in the wound and downregulates the plasma VEGF, NO, and SCF levels.

Wound closure lagged after Ad-PEDF injection compared with wound closure after control Ad-GFP injection (Supplementary Fig. 5A). X-gal staining showed that Wnt signaling was activated in the periphery of wounded skin (Fig. 6A and B), and the Ad-PEDF treatment reduced X-gal staining in the periphery of the wounded skin, compared with Ad-GFP treatment (Fig. 6C). Western blot analysis showed that PEDF reduced levels of p-LRP6 and levels of non-p-β-catenin in the wound tissue at days 4 and 7 after wounding, suggesting an inhibitory effect on Wnt signaling (Fig. 6D–I). Furthermore, p-LRP6 and non-p-β-catenin levels were increased in the wounded skin of PEDF KO mice compared with that in the WT mice (Fig. 6J–L). Our results showed that PEDF inhibits Wnt/β-catenin signaling in wound healing.

To further confirm that PEDF has a regulatory effect on wound healing through Wnt signaling, LiCl was injected into the 3-month-old diabetic mice. LiCl injection enhanced wound healing in diabetic mice, compared with that in the NaCl group (Supplementary Fig. 5B). The plasma VEGF levels and CD34⁺/Flk-1⁺ and CD133⁺/Flk-1⁺ cells were increased by LiCl injection compared with the NaCl controls (Fig. 7A–F). Western blot analysis showed that levels of non-p-β-catenin, VEGF, and SDF-1 levels were also increased in the LiCl injection group compared with the NaCl control group (Fig. 7G–J). These findings demonstrated that LiCl reached the wound bed and induced Wnt activation downstream of LRP6, the target of PEDF. Consequently, LiCl offset the PEDF-induced wound-healing delay. Taken together, these results suggest that the effect of PEDF on wound healing occurs, at least in part, through the blockade of Wnt signaling.

Impaired wound healing is a common cause of amputation in diabetic patients (1). The molecular basis underlying the pathogenesis of diabetes-induced wound-healing deficiency is not completely understood. The current study, for the first time, demonstrated that elevated PEDF levels in the plasma are associated with DFU. Consistent with the situation in human subjects, plasma PEDF levels are upregulated in diabetic mice. Using a PEDF Ab in diabetic mice and PEDF KO mice and the injection of recombinant PEDF, we have demonstrated that high levels of PEDF contribute to a delay in wound healing in diabetes through antiangiogenic activity. Using Wnt reporter mice and activation of Wnt signaling downstream of LRP6, our studies suggest that the effect of PEDF on wound healing occurs through blockade of the Wnt pathway. These observations established for the first time the association of elevated PEDF levels in diabetes with impaired wound healing and DFU, suggesting that PEDF is a promising drug target for the treatment of DFU.

PEDF is a multifunctional protein that is characterized by its broad functions, including antitumorigenicity, antimetastaticity (41,42), anti-inflammation (43), and antiangiogenicity (16) activities. Our previous study reported that circulation PEDF levels are elevated in patients with type 2 diabetes (21) and in patients with type 1 diabetes who have microvascular complications (22), while PEDF levels were decreased in renal (44) and retinal tissues (45). Here we showed the presence of elevated circulating PEDF levels in DM patients compared with DM+DFU patients and NDM individuals. Furthermore, our results showed that the plasma PEDF levels were increased in diabetic mice. This is the first evidence suggesting that high levels of PEDF may play a role in the regulation of wound healing in diabetes. To establish the role of PEDF in wound healing, we used a loss-of-function approach by using a PEDF Ab to block PEDF activity in diabetic mice and PEDF KO mice. Our results showed that the PEDF Ab not only improved delayed skin wound closure, but also increased MVD in the wound tissue. Diabetic patients have thinner skin and a reduced density of capillaries in the dermal layer, which results in higher morbidity from poorer blood flow response, neuropathy, ulceration, and gangrene, and can eventually develop into lower-limb amputation (46–48). The relationship between PEDF levels and these clinical characteristics is not yet entirely clear. Thus, the effect of PEDF on the skin, particularly on the reduction of vascular density in the panniculus adiposus layer, provides a possible explanation of how high levels of PEDF may contribute to impaired skin function and repair in patients with or at a high risk of peripheral vascular disease and amputation. In addition, patients with diabetic peripheral vascular disease often have impaired skin function and blood flow in lower-limb skin. Our results showed that PEDF KO mice had thicker resting skin from the top of the epidermis to the bottom of the dermis, an increased skin microvascular density, and higher blood flow rate than WT mice. In the gain-of-function approach, we injected recombinant PEDF into mice during wound healing. PEDF protein alone is sufficient to result in wound-healing delay in mice. Taken together, these approaches all suggest that elevated PEDF levels in the circulation indeed contribute to wound-healing delay in diabetes.

Both angiogenesis and vasculogenesis contribute to the process of wound healing. Recent evidence (49) indicated that vasculogenesis occurs during both of physiological development of vasculature and pathological neovascularization with postnatal vasculogenesis. Thus, it is a new target for pathological neovascularization through improvement of the decreased numbers and impaired function of EPCs. Our study showed that an intraperitoneal injection of the PEDF Ab in diabetic mice increased skin microvascular density, and improved the numbers and function of EPC mobilization and recruitment. The results of unchanged EPC abundance in bone marrow in diabetic mice suggest that PEDF might not affect the generation of EPCs.

Wnt signaling is known to play a key role in modulating angiogenesis (27,28) and is crucial for the regulation of EPC mobilization (33). However, the process of vasculogenesis has not been well understood. The numbers and functions of EPCs are crucial in the process of mobilization and recruitment. Our results demonstrated that decreased circulation numbers and dysfunction of EPCs correlate with Wnt signaling inhibition in the wound bed after injections of recombinant PEDF. Our previous study (24) showed that binding of PEDF to LRP6 blocked the activation of the Wnt pathway through hindering the Wnt ligand–induced dimerization of LRP6 and Frizzled receptor. To further confirm that Wnt signaling contributes to the mobilization of EPCs, LiCl, a potent activator of canonical Wnt signaling by inhibiting GSK-3β and stabilizing β-catenin, and subsequently activating VEGF expression (50,51), was used as an agent to bypass the effects of PEDF on Wnt signaling. The results demonstrated that LiCl inhibited the effects of PEDF on the mobilization of EPCs and β-catenin activation in wound healing. Taken together, our data suggest that PEDF is a Wnt/β-catenin inhibitor in postnatal murine skin. In addition, this study provides strong evidence that the process of mobilization and recruitment of bone marrow–derived EPCs contributes to the skin neovascularization in the wound-healing model, and that PEDF regulates EPC mobilization through the Wnt signaling pathway.

This EPC mobilization cascade starts with peripheral hypoxia-induced tissue release of VEGF (10,11). In this process, the increase of NO and SCF levels in the peripheral blood results in the mobilization of EPCs from bone marrow niches to circulation, ultimately allowing for their participation in vasculogenesis in wound tissue and wound healing (15). In the wounded tissue, EPC recruitment depends on ischemia-induced upregulation of SDF-1 (52). Because the mobilization and function of EPCs are regulated by VEGF, NO, and SCF, and the recruitment of EPCs is modulated by SDF-1, we measured and found increased expression of VEGF, NO, SCF, and SDF-1 after treatment with the PEDF Ab. Therefore, these observations strongly suggest that the PEDF Ab improved vascularization in the wound tissue by targeting EPCs. Although representative endogenous antiangiogenic proteins such as PEDF undoubtedly play a role in angiogenesis in vitro and in vivo (16), the regulatory role of PEDF in vasculogenesis has not been well established. Thus, our data demonstrate that PEDF blockage increases the numbers and functions of EPCs during the pathogenesis of wound neovascularization in a diabetic model.

Based on our results, we propose the following model: high levels of PEDF in diabetes inhibit Wnt/β-catenin signaling, leading to suppressed mobilization and function of EPCs, and, consequently, to angiogenesis and vasculogenesis during wound healing. These activities contribute to wound-healing delay in diabetic patients. Hence, blockade of PEDF may benefit the treatment of DFU and the prevention of amputation.

Acknowledgments. The authors thank Jim Henthorn in the flow cytometry core; Dr. Blake Hopiavuori for help with the use and analysis of laser Doppler flowmetry; and Dr. Yanhua Du for assistance with the endothelial progenitor cell culture.

Funding. This study was supported by National Nature Science Foundation of China grants 81172163, 81272338, 81272515, 81200706, and 81471033; National Key Sci-Tech Special Project of China grant 2013ZX09102-053; Program for Doctoral Station in University grants 20120171110053 and 20130171110053; Key Project of Nature Science Foundation of Guangdong Province, People’s Republic of China grant 10251008901000009; Key Sci-Tech Research Project of Guangdong Province, People’s Republic of China, grant 2011B031200006; Guandong Natural Science Fund grants S2012010009250, S2012040006986, and S2013010012520; Key Sci-Tech Research Project of Guangzhou Municipality, People’s Republic of China, grants 2011Y1-00017-8 and 12A52061519; and Changjiang Scholars and Innovative Research Team in University grant 985 Project PCSIRT 0947.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. W.Q. and C.Y. researched the literature, designed the methods and experiments, performed the laboratory experiments, acquired and analyzed the data, performed the statistical analysis, interpreted the results, wrote and approved the manuscript, and collaborated with all other authors. Z.D., D.C., J.F., Y.M., and X.G. codesigned the animal experiments, approved the manuscript, and collaborated with all other authors. R.C. and T.Z. worked together on the collection of associated data and their interpretation, approved the manuscript, and collaborated with all other authors. Z.W. and X.H. worked together on the isolation, culture, and associated assessment of endothelial progenitor cells, approved the manuscript, and collaborated with all other authors. L.Y., X.Y., and G.G. helped to define the study concepts; codesign the experiments; discussed analyses and their interpretation; helped to edit, revise, present, and approve the manuscript; and collaborated with all other authors. J.-x.M. codesigned the experiments; discussed analyses and their interpretation; helped to edit, revise, present, and approve the manuscript; and collaborated with all other authors. G.G. and J.-x.M. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.