Vascular endothelial growth factor stimulates bone repair by promoting angiogenesis and bone turnover John Street*†, Min Bao†‡, Leo deGuzman§, Stuart Bunting§, Franklin V. Peale, Jr.¶, Napoleone Ferrara‡, Hope Steinmetz§, John Hoeffel§, Jeffrey L. Cleland储, Ann Daugherty储, Nicholas van Bruggen§, H. Paul Redmond*, Richard A. D. Carano§, and Ellen H. Filvaroff‡** Departments of ‡Molecular Oncology, §Physiology, ¶Research Pathology, 储Pharmaceutical Research & Development, Genentech, Inc., 1 DNA Way, South San Francisco, CA 94080; and *Department of Academic Surgery, National University of Ireland, University College Cork, Cork, Ireland Communicated by Richard H. Scheller, Genentech, Inc., South San Francisco, CA, May 30, 2002 (received for review March 11, 2002)
Several growth factors are expressed in distinct temporal and spatial patterns during fracture repair. Of these, vascular endothelial growth factor, VEGF, is of particular interest because of its ability to induce neovascularization (angiogenesis). To determine whether VEGF is required for bone repair, we inhibited VEGF activity during secondary bone healing via a cartilage intermediate (endochondral ossification) and during direct bone repair (intramembranous ossification) in a novel mouse model. Treatment of mice with a soluble, neutralizing VEGF receptor decreased angiogenesis, bone formation, and callus mineralization in femoral fractures. Inhibition of VEGF also dramatically inhibited healing of a tibial cortical bone defect, consistent with our discovery of a direct autocrine role for VEGF in osteoblast differentiation. In separate experiments, exogenous VEGF enhanced blood vessel formation, ossification, and new bone (callus) maturation in mouse femur fractures, and promoted bony bridging of a rabbit radius segmental gap defect. Our results at specific time points during the course of healing underscore the role of VEGF in endochondral vs. intramembranous ossification, as well as skeletal development vs. bone repair. The responses to exogenous VEGF observed in two distinct model systems and species indicate that a slow-release formulation of VEGF, applied locally at the site of bone damage, may prove to be an effective therapy to promote human bone repair.
B
one repair is a multistep process involving migration, proliferation, differentiation, and activation of several cell types (1, 2). Bone formation can occur through two distinct processes. If bone segments are stabilized, or during development of some skull and facial bones, mesenchymal precursor cells differentiate directly into bone-forming osteoblasts in a process called intramembranous ossification. Alternatively, in a biomechanically unstable environment, or in development of long bones and vertebrae, bone formation occurs via a cartilage intermediate in a process called endochondral ossification (1, 2). Expression of particular growth factors—such as fibroblast growth factors (FGFs), platelet-derived growth factors (PDGFs), transforming growth factor-betas (TGF-s), vascular endothelial growth factor (VEGF), and bone morphogenetic proteins (BMPs)—during the course of healing suggests a possible role for these secreted factors in bone repair. In fact, each of these factors, except VEGF, has been shown to stimulate bone healing in animal models (1, 2). Although VEGF can control hypertrophic cartilage structure and vascularity within the developing growth plate (3), the role of VEGF in bone repair has not yet been determined. VEGF is expressed in the fracture callus in animal models in much the same temporal and spatial pattern as during long bone development (4, 5). Other pro- and antiangiogenic factors expressed in the growth plate of developing bones (6) are also present in the fracture callus during repair (4, 5, 7). Thus, the fracture callus contains many factors
9656 –9661 兩 PNAS 兩 July 23, 2002 兩 vol. 99 兩 no. 15
that could promote bone healing by coordinating angiogenesis with bone homeostasis (8). To elucidate the role of endogenous VEGF in vivo, we examined the activity of a soluble, neutralizing VEGF receptor, Flt-IgG (9), in models of endochondral ossification (femoral fractures) and intramembranous ossification (tibial cortical bone defects). Inhibition of VEGF had distinct effects on bone healing in each model, as assessed by histology, quantitation of vascularity, and computed tomography (CT), a predictor of mechanical stability of healing bones (10). Taken together with data presented herein that VEGF directly promotes the differentiation of primary osteoblasts in vitro, we conclude that at least part of the mechanism(s) by which VEGF inhibition impaired bone repair is through disruption of osteoblast differentiation. Our results with VEGF in a murine model of fracture repair and a rabbit critical size defect model suggest that exogenous VEGF could enhance bone repair in a spectrum of orthopedic conditions. Our data show distinct roles for VEGF in endochondral vs. intramembranous bone repair and also provide insight into how VEGF might be used to promote healing of bone defects. Materials and Methods Animals and Reagents. Six- to 8-wk-old male C57BL6 mice
(Charles River Laboratories) and 3.8- to 4.2-kg male New Zealand White (NZW) Rabbits (Western Oregon Rabbitry, Philomath, OR) were used in accordance with National Guidelines.
Murine Femoral Fracture Healing Model. A midshaft, fixed femur fracture was created in anesthetized mice (n ⫽ 175) (11). To create a ‘‘challenged’’ fracture, the periosteum, a region critical for healing (12), was stripped for 2.0 mm proximal and distal to the fracture site. Ten microliters of the polylactic acid depot formulation, PLAD (see below) ⫾ VEGF (10 g) was applied at the fracture site. Any animals in which the pin came out, the fracture was grossly displaced, or the fracture was not midshaft (as assessed by radiographs) were not analyzed. With these inclusion criteria, experiments had a minimum of seven animals per group. Creation of Focal Cortical Defect in the Tibia of Mice. A full thickness
unicortical defect was created on the anteriomedial aspect of the right tibia using a dental burr (1 mm), with continuous saline
Abbreviations: VEGF, vascular endothelial growth factor; CT, microcomputed tomography; bFGF, basic fibroblast growth factor; PECAM, platelet-endothelial cell adhesion molecule; BMP, bone morphogenetic protein; VOI, volume of interest; PLAD, polylactic acid depot; TGF-, transforming growth factor-beta. †J.S.
and M.B. contributed equally to this work.
**To whom reprint requests should be addressed. E-mail:
[email protected].
www.pnas.org兾cgi兾doi兾10.1073兾pnas.152324099
In Vivo Platelet-Endothelial Cell Adhesion Molecule (PECAM)-1 Labeling. Monoclonal antibody (mAb) rat anti-mouse PECAM-1
IgG2a (PharMingen, clone MEC13.3) labeled with 125I (Dupont NEN, Boston, MA, NEZ-033A) and a nonspecific isotype control antibody, rat anti-mouse CD35, IgG2a (PharMingen Inc., San Diego, CA, clone 8C12) labeled with 131I (Dupont NEN, NEZ035A) were used (14–16). All antibodies were iodinated using the iodogen method in a ratio of 1 g of antibody to 1 Ci (1 Ci ⫽ 37 GBq) of either 125I or 131I (15). To measure PECAM-1 binding, a mixture of 125I PECAM-1 mAb (10 g) and 131I nonspecific mAb (equivalent to 500,000 cpm) was diluted with PBS (to 200 l). Cold PECAM-1 mAb (30 g) was added, and the mixture was injected through the jugular vein catheter and allowed to circulate for 5 min (16). Blood samples were obtained from carotid catheters to measure circulating radiolabeled antibody levels. Animals were exsanguinated by perfusion with 6 ml bicarbonate buffered saline (BBS) through the jugular catheter with simultaneous blood withdrawal from the carotid catheter. This procedure was followed by perfusion of BBS through the carotid catheter (15 ml) after severing the inferior vena cava at the thoracic level. Organs and muscles were collected and weighed, and radioactivity was counted. Formulation of VEGF. Bioerodible polylactic acid depot (PLAD)
uses PLA (Resomer R 202 H, Boehringer Ingleheim), with hydrophilic (benzyl alcohol, BA) and hydrophobic (benzyl benzoate, BB) solvents, to dissolve protein and solvate PLA (17). Liquid recombinant murine VEGF (muVEGF; Genentech) was spray freeze-dried to produce a powder for formulation as a solid phase in PLAD. PLAD solutions were prepared from a mixture of 40% (wt兾wt) PLA, 5% (wt兾wt) BA, (low peroxide, double distilled USP grade, Genentech) and 55% (wt兾wt) BB (USP grade; Sigma) (17). PLAD muVEGF was made by homogenizing (5 mm microfine shear homogenizer; VirTis) muVEGF powder with this PLAD solution for 2 min at 8,000 rpm. Street et al.
from the radius 1.2 cm along the mid-shaft of 30 anesthetized, male NZW rabbits. One centimeter of the radius was excised by using a sterile saw blade with liberal saline irrigation to prevent overheating of bone margins. A local, subcutaneous osmotic pump [Alzet (Palo Alto, CA) model 2001, 1 l兾h] was used to deliver VEGF (0, 50, 100, 250, and 1,000 g; six animals per group) for 7 days. Analgesics were given before surgery and for 72 h postsurgery. Primary Human Osteoblast Cultures. After extensive washes of trabecular bone explants from consenting young adults with no evidence of metabolic bone disease, small bone chips (1 ⫻ 1 ⫻ 1 mm) were placed in flasks with ␣-modified Earle’s medium, 10% heat-inactivated FCS, penicillin (100 units兾ml), and streptomycin (50 g兾ml) and cultured at 37°C and 5% CO2. Fluorescence-activated cell sorter analysis for osteocalcin indicated ⬇97% cell purity. Approximately 98% of cells bound biotinylated VEGF. Experiments were performed on osteoblasts subcultured to passage 3–6. For differentiation assays, osteoblasts were seeded in six-well plates. Upon confluence, 50 g兾ml ascorbic acid and recombinant human VEGF 165 (0–50 ng兾ml) or a monoclonal mouse anti-human VEGF neutralizing antibody (0.3 g兾ml) were added. Media were replenished daily. When mineralized nodules began to appear (3–8 days), beta-glycerolphosphate was added. Mineral (von Kossa) staining was performed at 18 days. All cultures were analyzed in triplicate, and six fields per culture well were counted. Alkaline phosphatase activity was measured in cell supernatants (Sigma Diagnostics). For VEGF assays, primary human osteoblasts were plated (1 ⫻ 105 cells per well of 24-well plates) for 6–8 h until adherence. After changing media, cells were cultured for 6, 12, 18, 24, 36, or 48 h. Cell supernatants were centrifuged at 3,000 ⫻ g for 10 minutes at 4°C, 0.2 m sterile filtered, and assayed for human VEGF and bFGF by an ELISA (Quantikine Immunoassay, R&D Systems, Minneapolis).
Results Endogenous VEGF Is Essential for Normal Fracture Repair. Treatment
of mice with a soluble VEGF receptor, Flt-IgG, during the course of fracture repair, dramatically impaired new bone formation (Fig. 1A). Repair was examined at time points corresponding to soft (cartilaginous) callus (7 days) and hard (bony) callus (14 days) (4, 5). Quantitation by CT indicated that volumes of both the total and calcified callus were decreased by 35.5% (P ⫽ 0.03) and 43.9% (P ⫽ 0.01), respectively, in Flt-IgG-treated mice at 7 days, but not at 14 days after fracture relative to mice treated with control antibodies (Table 1, which is published as supporting information on the PNAS web site, www.pnas.org; and data not shown). However, percent callus calcified was decreased at both time points, i.e., by 10.6% (P ⫽ 0.04) at 7 days (Fig. 1B) and by 19.4% (P ⫽ 0.008) at 14 days (Fig. 1C). Similarly, mineral density was decreased by 10.7% in the total callus (P ⫽ 0.005) and 6.4% in the calcified callus (P ⫽ 0.0005) at 7 days (Fig. 1D), and 13.4% in the total callus (P ⫽ 0.005) and 5.7% in the calcified callus (P ⫽ 0.02) at 14 days (Fig. 1E). Histologic examination of the fracture callus at 7 days postfracture (Fig. 1 F–H) showed wider callus trabeculae and persistent perinuclear osteocyte lacunae in the fracture callus of Flt-IgG-treated mice (Fig. 1G), suggesting decreased callus maturity relative to both sets of control mice (Fig. 1 F and H). Endogenous VEGF Is Required for Normal Repair of Cortical Bone Defects. Intramembranous healing of a focal defect in the tibial
cortex was also impaired in Flt-IgG-treated mice (Fig. 2A). Relative to control antibody-treated animals, total and calcified callus volumes were reduced by 17.3% (P ⫽ 0.035) and 60.7% (P ⫽ 4.5 ⫻ 10⫺7), respectively, at 7 days (Table 1). Percent PNAS 兩 July 23, 2002 兩 vol. 99 兩 no. 15 兩 9657
SCIENCES
CT Analysis. X-ray microcomputed tomography (CT) images were acquired at 50 kV and 80 (mice) or 160 (rabbits) microamperes (A) by using a CT20兾40 (SCANCO Medical, Bassersdorf, Switzerland). Axial images were obtained [26 ⫻ 26 ⫻ 35 m, and an inter-slice gap of 69 m (mice), 30 ⫻ 30 ⫻ 31 m and contiguous slices (rabbits)]. A hydroxyapatite phantom (2.91 g兾cm3) was used for system calibration. Callus volume and mean voxel intensity were calculated for a callus volume of interest (VOIcallus). A ‘‘calcification’’ threshold (0.48 gHA兾cm3), which equals 50% of the minimum intensity required to segment cortical bone, was applied to VOIcallus to determine volume and mean intensity of calcified callus. Percent calcified callus was defined as the ratio of calcified callus volume to total callus volume. VOIcallus for mouse bones was determined manually using SCANCO image analysis software. VOIcallus for rabbit bones was determined with an in-house segmentation algorithm developed with ANALYZE software (AnalyzeDirect, Lenexa, KS). Lower and upper thresholds (0.22 and 1.33 gHA兾cm3, respectively), determined by histogram analysis of data from three rabbits, were applied to extract potential callus voxels. A series of morphological filtering (13) operations (erode, open, conditional dilate, and close) were applied to extract the callus volume. Algorithm results were highly correlated (r ⱖ 0.96, P ⬍ 0.01) with manual estimates for an independent group of eight rabbit bones.
Rabbit Radius Segmental Gap Model. The periosteum was removed
APPLIED BIOLOGICAL
irrigation to prevent thermal necrosis of margins. Mice were untreated (Control) or were given i.p. injections (25 mg兾kg) of a control IgG (anti-glycoprotein D) or murine Flt(1–3)-IgG (9) on alternate days.
Fig. 1. Flt-IgG treatment impairs fracture repair. (A) Three-dimensional rendering of femurs from CT images of Control or Flt-IgG- or IgG-treated mice 7 days after fracture. Pseudocoloring was applied to highlight callus (blue) and cortical bone (peach). (B and C) Percent calcified callus for control (Con) or Flt-IgG (Flt)or IgG (IgG)-treated mice at (B) 7 or (C) 14 days. Mean mineral density of Total or Calcified callus at (D) 7 or (E) 14 days. Shown are means ⫾ SEM. *, P ⬍ 0.05; **, P ⬍ 0.01, relative to IgG-treated mice. (F–H) Hematoxylin兾eosin (H&E) staining of the fracture site at 7 days in (F) Control or (G) Flt-IgG- or (H) IgG-treated mice. Note wider and less mineralized trabeculae and large lacunae around osteocytes in (G) Flt-IgG-treated mice relative to (F and H) controls.
calcification was 50.7% lower (P ⫽ 3.04 ⫻ 10⫺7) at 7 days (Fig. 2B), but only trended lower (P ⫽ 0.095) at 14 days (Fig. 2C). Treatment with Flt-IgG decreased 7-day total callus mean mineral density by 24.6% (P ⫽ 1.11 ⫻ 10⫺5) and calcified mean mineral density by 7.9% (P ⫽ 0.004) (Fig. 2D) and 14-day total callus mean mineral density by 17.5% (P ⫽ 0.003) and calcified callus mean mineral density by 14% (P ⫽ 0.0003) (Fig. 2E). In contrast to the well-mineralized, woven bone in defect sites in control animals at 7 days after surgery, Flt-IgG-treated animals showed largely unresorbed, uncalcified hematoma in the medullary space (Fig. 2 F–I). At 14 days, mineralized bone was present at the defect site of all mice, but control animals showed evidence of remodeling to lamellar bone, whereas in Flt-IgG treated animals, defects were still largely composed of woven bone with less well-organized collagen fibrils when viewed in polarized light (data not shown), suggesting that VEGF affects bone cell activity. VEGF Directly Enhances Osteoblast Activity in Vitro. To test the
possibility that VEGF had direct effects on osteoblasts, we analyzed primary human osteoblasts in vitro. VEGF increased nodule formation and alkaline phosphatase activity (18) in a dose-dependent manner (Fig. 3 A and B). Unlike a mouse preosteoblast cell line (19), primary human osteoblasts responded to inhibition of VEGF with a 43.3% decrease in nodule
9658 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.152324099
formation and a 39.3% decline in alkaline phosphatase production (Fig. 3 A and B). Under hypoxic conditions, VEGF (20), but not bFGF, was up-regulated by as much as 96.6% (Fig. 3 C and D). Thus, hypoxia-induced up-regulation of VEGF in osteoblasts at the site of injury may contribute to local osteoblast activation and promote initial calcification of the fracture hematoma. Endogenous VEGF Promotes Callus Vascularity. Vascularity, as quan-
titated by PECAM expression (14, 15) was reduced by 18% (P ⫽ 0.01) in fractured bones of Flt-IgG-treated animals relative to IgG-treated mice at 7 days (Fig. 4A). Because of injury, PECAM expression in soft tissues of fractured (right) legs was increased relative to contralateral, intact (left) legs in all groups (Fig. 4B). However, percent induction of vascularity in the fractured leg relative to the untreated, contralateral leg was ⬇22.5% less (P ⬍ 0.05) in Flt-IgG-treated mice relative to IgG or untreated control mice (Fig. 4C). No significant difference in vascular induction between Control and IgG-treated mice was found (Fig. 4C).
Local VEGF Treatment Stimulates Bone Repair. To test the effect of
VEGF on repair of femur fractures, we created a challenged fracture by disrupting the periosteum (12). Treatment with slow-release VEGF (10 g) caused a 33.6% increase (P ⫽ 0.0001) in percent calcified callus (Fig. 5A), and augmented mean mineral density in the total callus by 19.2% (P ⫽ 0.0001)
Street et al.
and calcified callus by 5.2% (P ⫽ 0.007) (Fig. 5B). As in previous experiments (Fig. 4B), vascularity was increased in soft tissues in both the upper and lower legs on the fractured side (right) relative to that of the contralateral, intact (left) legs in control mice (Fig. 5 C and D). Vascularity in the upper leg (the site of VEGF application) (Fig. 5C), but not the lower leg (Fig. 5D), of VEGF-treated femur fractures was 26% (P ⫽ 0.0002) higher than that of carrier-treated femur fractures (Fig. 5C). Thus, although femur fractures caused increased vascularity in both the upper and lower part of the injured leg, VEGF treatment in the femur led to local enhancement of angiogenesis in the upper, but not the lower, leg. VEGF Stimulates Repair of a Critical Size Defect in Rabbit Radii. To test the therapeutic potential of VEGF in another species and model, we created segmental defects in rabbit radii and implanted a pump with various doses of VEGF (0, 50, 100, 250, and 1,000 g) continuously released over the first 7 days after surgery. Consistent with the critical size (10 mm) of these defects, placebotreated defects were not able to create a bony bridge across the gap (Fig. 6A). In contrast, VEGF treatment caused significant filling with bone (Fig. 6B). For example, VEGF at 250 g, caused a 91% increase (P ⫽ 0.02) in total callus volume and a 95% increase (P ⫽ 0.02) in calcified callus volume (Fig. 6C).
Discussion VEGF inhibition in mice disrupted repair of femoral fractures and cortical bone defects. Our data provide evidence that VEGF Street et al.
activity is essential for appropriate callus formation and mineralization in response to bone injury. Fracture repair occurs in a series of steps, involving an initial inflammatory phase, a soft callus (cartilage) phase, a hard callus (bone) phase, and a remodeling phase (1, 2, 4, 5). Consistent with the fact that endogenous VEGF is expressed in hypertrophic chondroctyes at ⬇10 days after fracture in mice (4, 5), Flt-IgG affected calcification of the hard callus 14 days after fracture. Our data show that VEGF is involved in conversion of the soft, cartilaginous callus to a hard, bony callus during fracture repair just as VEGF couples angiogenesis, cartilage resorption, and ossification in the growth plate of developing mice (3). Our studies of vascularity at 7 days after fracture indicate the importance of VEGF in early angiogenesis. In the early stages of bone repair, large amounts of active VEGF are found in the fracture hematoma (21), a VEGF source that is not present in developing bones. Thus, Flt-IgG inhibition of soft callus formation at 7 days after fracture could not have been predicted based on developmental studies. Our results show a role for VEGF in the early stages of fracture repair and highlight differences in the processes of endochondral ossification during development vs. that in fracture repair. As in the fracture model, Flt-IgG treatment impaired healing in the cortical bone defect model. The persistence of unresorbed, unmineralized fracture hematoma in this defect at 7 days and the persistence of woven bone at 14 days indicate remodeling defects in Flt-IgG-treated mice. Such results are consistent with previous findings that VEGF stimulates chemotaxis of osteoclasts PNAS 兩 July 23, 2002 兩 vol. 99 兩 no. 15 兩 9659
SCIENCES
APPLIED BIOLOGICAL
Fig. 2. Flt-IgG treatment impairs cortical bone defect repair. (A) Three-dimensional rendering of tibiae from CT images of Control or Flt-IgG- or IgG-treated mice 7 days after surgery. (B and C) Percent calcified callus of Control (Con) or Flt-IgG (Flt)- or IgG (IgG)-treated mice at (B) 7 or (C) 14 days. Mean mineral density of Total or Calcified callus at (D) 7 or (E) 14 days. Shown are means ⫾ SEM. **, P ⬍ 0.01 relative to IgG-treated mice. H&E staining of tibiae at 7 days in (F) adjacent tissue or the defect site in (G) Control or (H) Flt-IgG- or (I) IgG-treated mice. Little evidence of resorption or mineralization of red cells masses in (H) Flt-IgG-treated mice was in contrast to (G and I) control animals, in which the medullary callus contained trabeculae with well-mineralized woven bone.
Fig. 3. VEGF initiates a positive autocrine loop in primary human osteoblasts. (A) Nodule formation or (B) alkaline phosphatase (AP) activity after treatment with VEGF (0, 1, 5, 10, 25, or 50 ng兾ml) or anti-VEGF (␣V) (0.3 g兾ml) (C) VEGF and (D) bFGF production under 21% O2, i.e., normoxic (N) or 2% O2, i.e., hypoxic (H) conditions for 6 (6h), 12 (12h), 24 (24h), 36 (36h), or 48 (48h) hours after cell adherence. Shown are means ⫾ SEM. For A and B, *, P ⬍ 0.05, **, P ⬍ 0.01 vs. control (0). For C, *, P ⬍ 0.05 and **, P ⬍ 0.01 for hypoxic vs. normoxic samples.
(22) and osteoclast activity (23). Unlike the timing of Flt-IgG effects on endochondral bone formation (fractures), the effects of Flt-IgG on intramembranous bone formation (cortical defects) were more prominent at the earlier (7 day) than later (14 day) time point. The difference in the effect of Flt-IgG at later time points in the two models is related to the fact that intramembranous ossification does not involve a cartilage intermediate. Thus, the mechanism of Flt-IgG inhibition of repair of the cortical bone defect at 14 days is distinct from that in the
Fig. 4. Flt-IgG treatment decreases angiogenesis. PECAM measurements of (A) bone (femur) from the fractured leg (Right) and the contralateral control leg (Left), or (B) tissue from the left (L) normal and right (R) fractured leg of control (Con) or Flt-IgG (Flt)- or IgG (IgG)-treated mice at 7 days. Shown are means ⫾ SEM. **, P ⬍ 0.01 for fractured legs (Right) relative to the contralateral (Left) legs within each treatment group. (C) Percent increase [(PECAMright ⫺ PECAMleft)兾PECAMleft] in vascularity in the fractured leg relative to the contralateral control leg. In A and C, *, P ⬍ 0.05 for Flt-IgG vs. both control and IgG-treated mice. 9660 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.152324099
Fig. 5. Local VEGF treatment promotes fracture repair. (A) Percent calcified callus in mice treated with VEGF (10 g) (V) or carrier alone (⫺). (B) Mean mineral density of the Total or Calcified callus of mice (C and D). PECAM measurements in (C) upper legs or (D) lower legs of bones [Femur (C) or Tibia (D)] or soft tissues (Tissue) for the right, fractured leg (R) or the contralateral, left leg (L) at 7 days. Shown are means ⫾ SEM. *, P ⬍ 0.05, **, P ⬍ 0.01 for VEGF-treated (V) relative to carrier alone (⫺).
fracture model. In the absence of cartilage, osteoblasts are likely producing, and responding to, VEGF as indicated by our in vitro data showing hypoxia-induced expression of VEGF, but not
Fig. 6. VEGF stimulates repair of a segmental gap defect. Three-dimensional (3D) renderings of CT images of radius critical defects at 28 days in (A) Placebo and (B) VEGF (250 g)-treated rabbits. The gap remaining in the VEGF-treated bone is the site of the pump catheter. (C) Volumes of the Total Callus or Calcified Callus. Shown are means ⫾ SEM. *, P ⬍ 0.05, **, P ⬍ 0.01 relative to control (0).
Street et al.
1. Mandracchia, V. J., Nelson, S. C. & Barp, E. A. (2001) Clin. Pod. Med. Surg. 18, 55–77. 2. Gittens, S. A. & Uludag, H. (2001) J. Drug Targeting 9, 407–429. 3. Gerber, H. P., Vu, T. H., Ryan, A. M., Kowalski, J., Werb, Z. & Ferrara, N. (1999) Nat. Med. 5, 623–628. 4. Ferguson, C., Alpern, E., Miclau, T. & Helms, J. A. (1999) Mech. Dev. 87, 57–66. 5. Tatsuyama, K., Maezawa, Y., Bab, H., Imamura, Y. & Fukuda, M. (2000) Eur. J. Histochem. 44, 269–278. 6. Gerber, H. P. & Ferrara, N. (2000) Trends Cardiovasc. Med. 10, 223–228. 7. Hadjiargyrou, M., Ahrens, W. & Rubin, C. T. (2000) J. Bone Miner. Res. 15, 1014–1023. 8. Glowacki, J. (1998) Clin. Orthop. 355, S82–S89. 9. Ferrara, N., Chen, H., Davis-Smyth, T., Gerber, H. P., Nguyen, T. N., Peers, D., Chisholm, V., Hillan, K. J. & Schwall, R. H. (1998) Nat. Med. 4, 336–340. 10. Augat, P., Merk, J., Genant, H. K. & Claes, L. (1997) Calcif. Tissue Int. 60, 194–199. 11. Bonnarens, F. & Einhorn, T. A. (1984) J. Orthop. Res. 2, 97–101. 12. Utvag, S. E., Grundnes, O. & Reikeraos, O. (1996) J. Orthopaed. Trauma 10, 279–284. 13. Jain, A. K. (1989) in Fundamentals of Digital Image Processing, ed. Kailath, T. (Prentice–Hall, Englewood Cliffs, NJ), pp. 384–389. 14. Vecchi, A., Garlanda, C., Lampugnani, M. G., Resnati, M., Matteucci, C., Stoppacciaro, A., Schnurch, H., Risau, W., Ruco, L., Mantovani, A., Dejana, E. (1994) Eur. J. Cell. Biol. 63, 247–254. 15. Eppihimer, M. J., Russell, J., Langley, R., Vallien, G., Anderson, D. C. & Granger, D. N. (1998) Microcirculation 5, 179–188. 16. Panes, J., Perry, M. A., Anderson, D. C., Manning, A., Leone, B., Cepinskas, G., Rosenbloom, C. L., Miyasaka, M., Kvietys, P. R. & Granger, D. N. (1995) Am. J. Physiol. 269, H1955–H1964. 17. Cleland, J. L., Duenas, E. T., Park, A., Daugherty, A., Kahn, J., Kowalski, J. & Cuthbertson, A. (2001) J. Controlled Release 72, 13–24. 18. Midy, V. & Plouet, J. (1994) Biochem. Biophys. Res. Commun. 199, 380–386. 19. Deckers, M. M., Karperien, M., van de Bent, C., Yamashita, T., Papapoulos, S. E. & Lowik, C. W. (2000) Endocrinology 141, 1667–1674.
20. Steinbrech, D. S., Mehrara, B. J., Saadeh, P. B., Greenwald, J. A., Spector, J. A., Gittes, G. K. & Longaker, M. T. (2000) Am. J. Physiol. Cell. Physiol. 278, C853–C860. 21. Street, J., Winter, D., Wang, J. H., Wakai, A., McGuinness, A. & Redmond, H. P. (2000) Clin. Orthop. 378, 224–237. 22. Engsig, M. T., Chen, Q.-J., Vu, T. H., Pedersen, A.-C., Therkidsen, B., Lund, L. R., Henriksen, K., Lenhard, T., Foged, N. T., Werb, Z., et al. (2000) J. Cell. Biol. 151, 879–889. 23. Niida, S., Kaku, M., Amano, H., Yoshida, H., Kataoka, H., Nishikawa, S., Tanne, K., Maeda, N., Nishikawa, S. I. & Kodama, H. (1999) J. Exp. Med. 190, 293–298. 24. Mayr-Wohlfart, U., Waltenberger, J., Hausser, H., Kessler, S., Gunther, K. P., Dehio, C., Puhl, W. & Brenner, R. E. (2002) Bone 30, 472–477. 25. Khan, S. N., Bostrom, M. P. & Lane, J. M. (2000) Orthop. Clin. N. Am. 31, 375–388. 26. Mori, S., Yoshikawa, H., Hashimoto, J., Ueda, T., Funai, H., Kato, M. & Takaoka, K. (1998) Bone 22, 99–105. 27. Street, J., Wang, J. H., Power, C., Wakai, A. & Redmond, H. P. (2000) Surg. Forum 51, 465–467. 28. Hay, E., Lemonnier, J., Fromigue, O. & Marie, P. J. (2001) J. Biol. Chem. 276, 29028–29036. 29. Yamashita, H., Shimizu, A., Kato, M., Nishitoh, H., Ichijo, H., Hanyu, A., Morita, I., Kimura, M., Makishima, F. & Miyazono, K. (1997) Exp. Cell Res. 235, 218–226. 30. Claffey, K. P., Abrams, K., Shih, S. C., Brown, L. F., Mullen, A. & Keough, M. (2001) Lab. Invest. 81, 61–75. 31. Deckers, M. M., van Bezooijen, R. L., van der Horst, G., Hoogendam, J., van der Bent, C. Papapoulos, S. E. & Lowik, C. W. (2002) Endocrinology 143, 1545–1553. 32. Yeh, L. C. & Lee, J. C. (1999) Mol. Cell. Endocrinol. 153, 113–124. 33. Chua, C. C., Hamdy, R. C. & Chua, B. H. (2000) Biochim. Biophys. Acta 1497, 69–76. 34. Kozawa, O., Matsuno, H. & Uematsu, T. (2001) J. Cell. Biochem. 81, 430–436. 35. Goad, D. L., Rubin, J., Wang, H., Tashjian, A. H. & Patterson, C. (1996) Endocrinology 137, 2262–2268.
Street et al.
PNAS 兩 July 23, 2002 兩 vol. 99 兩 no. 15 兩 9661
SCIENCES
We thank Alisha Eisert, Julio Ramirez, Jed Ross, and David Wood for technical support.
APPLIED BIOLOGICAL
ability to couple angiogenesis with bone formation and remodeling. In addition, VEGF may act as a central mediator for other factors. Accordingly, inhibition of VEGF blocks the angiogenic activity of bFGF and BMP-2 (30, 31) and induction of osteoblast differentiation by OP-1 (32). Furthermore, most osteoinductive factors stimulate VEGF production (30–35). Thus, stimulation of bone repair by such factors in animal models or clinical trials may be through stimulation of VEGF production. Our results indicate that slow-release VEGF could be an effective therapeutic for bone injuries and suggest that a patient with bone damage who has alterations in VEGF regulation and兾or responsiveness may have a relatively poor prognosis. Given that significant cell death occurs in the first 24 h after fracture (21) and that shorter treatments with VEGF protein in vivo were less effective (data not shown), a slow-release formulation (such as that used in the mouse studies) may be necessary to promote optimal healing by VEGF. Such a VEGF formulation could also prove useful in additional indications such as spinal fusion, non-unions, and maxillo-facial surgeries.
bFGF, in primary osteoblasts. The decrease in mean mineral density at 14 days in the cortical defect might be explained by interference with stimulation of osteoblasts by VEGF as our in vitro data with anti-VEGF antibodies indicate. VEGF also acts to recruit and activate osteoclasts as well as stimulate osteoblast chemotaxis (24), differentiation, and matrix mineralization. Our results implicate VEGF in intramembranous ossification. The varied mechanism of action of growth factors that can promote repair in animal models (1, 2) illustrates how stimulation of different cellular activities can contribute to the multistep process of bone regeneration (1, 2, 25). For example, BMP-2 can convert mesenchymal cells to committed osteoblasts, with angiogenesis being central to its ability to induce ectopic bone (26). Although both VEGF and BMP-2 can stimulate osteoblast migration (18) and differentiation (18, 19), osteoblast apoptosis can be inhibited by VEGF (27) and promoted by BMP-2 (28). Although bFGF, TGF-, and some BMP family members (29) can promote blood vessel formation, bFGF promotes osteoblast proliferation, but not differentiation (2, 18), and the ability of TGF- to augment fracture healing is relatively weak (2, 25). Thus, one possible advantage of local VEGF therapy may be its